BenchmarkComparisons.md

BenchmarkComparisons

Live progress repair under ADR-085 owns bounded Actions visibility and cancellation diagnostics for native workloads.

Current complete-cohort publication

Owner-directed immediate legacy removal is governed by ADR-076. Its live path supersedes the historical comparison-smoke/comparison-suite and three-profile publication contracts below; immutable historical evidence keeps its original provenance. Source implementation and genuine runtime/publication qualification are distinct.

Requirement	Measurable acceptance	Tests and evidence
REQ-BC-CURRENT-001: the complete current native cohort is the sole live producer	AC-BC-CURRENT-001: exact current Benchmarks run/attempt/source, all 270 required cells and 277 original inputs pass existing correctness/fairness/provenance checks; missing, expired, mixed, skipped or failed inputs reject	SiteIsolatedGitHub selection/proof/archive/input/authority tests; actual aggregate and original GitHub archives
REQ-BC-CURRENT-002: retire superseded collector, archives and current-profile assets together	AC-BC-CURRENT-002: no active comparison-smoke/comparison-suite collector, twelve-file precondition or three-profile current renderer remains; replacement retains all applicable safety, accessibility, vendor, arithmetic and browser criteria	scoped dependency inventory, workflow source tests and standalone isolated site/browser tests
REQ-BC-CURRENT-003: preserve unchanged qualified authority around final build	AC-BC-CURRENT-003: bounded immutable receipt, original archive and all 277 input hashes match before/after builder; emitted aggregate bytes match; every remaining native coverage/no-skip/source-closure gate passes	SiteIsolatedGitHub integrity/build/coverage tests, original TRX/Cobertura/Node/browser artifacts and schema-v2 publication receipt
REQ-BC-CURRENT-004: deploy only fresh exact-source evidence	AC-BC-CURRENT-004: current main and exact original producer tuple match immediately before needs-gated least-privilege Pages delivery; altered source/attempt/artifact rejects and actual provider receipt records outcome	SiteIsolatedGitHubFreshnessTests, workflow source tests, genuine same-run qualification and Pages provider receipt

Canonical ownership remains BenchmarkComparisons under site/scripts/workflows and SiteTests; the precise disjoint tasks and integration points are in ADR-076.

Canonical engine decision and candidate cleanup

ADR-071 records the owner's ZoneTree/ZoneTree.FullTextSearch choice and immediate Garnet/Tsavorite removal.

Requirement	Acceptance	Mapping
REQ-ZT-001: active storage experiments use the selected canonical engine	AC-ZT-001: no Garnet package or Tsavorite implementation/dispatch remains; real ZoneTree raw/scaled correctness, full loaded-record counts, bounds and cleanup still hold; removed labels fail explicitly	existing RawStorage and ScaledRawStorage TUnit cases, dependency/source inspection
REQ-ZT-002: delete discarded plans and unused pipeline paths immediately	AC-ZT-002: active docs have no dead candidate-plan links; Benchmarks has no raw/internal-codec job or diagnostic-only bypass, and preserves every native comparison/image/aggregate/site gate	workflow contract TUnit cases, governance and root diff review

Immutable prior result files remain truthful historical evidence. This cleanup does not establish performance, RF3 durability or production readiness. Root owns package/workflow/docs joins; the bounded worker owns candidate fixture and matching report/test removal under ADR-071. UI N/A: internal diagnostics do not produce the public comparison website.

Native gate repair / ADR-068 preserves full isolated scope while repairing actual stale startup job proof and investigating Kurrent metadata/leader failures from dbd01269 original artifacts. The bounded same-ID current-job refresh and exact canonical fixture metadata contracts are frozen; native leader routing and complete cohort qualification remain open.

Shared parallel performance pipeline

REQ-PIPE-005/006 and AC-UB-001..006 in shared workflow acceptance under ADR-064 repair orchestration: common image preparation includes pinned TimeSeries checks; independent build, plan and images start in parallel; preflight, CRUD and specialized matrices share only required plan/image inputs and run in parallel without an arbitrary cap. Every authored job/step has a concrete readable name. Exact-source display-name validators and independent test oracles change together, preserving internal job IDs, artifact/schema/permission contracts and historical report bytes.

flowchart LR
    Plan[Plan benchmark runs] --> Preflight[Database checks]
    Plan --> Crud[Document benchmarks]
    Plan --> Models[Vector queue graph stream benchmarks]
    Images[Docker images and shared workload checks] --> Preflight
    Images --> Crud
    Images --> Models
    Build[Build and checks] --> Aggregate[Combine all results]
    Preflight --> Aggregate
    Crud --> Aggregate
    Models --> Aggregate
    Aggregate --> Website[Check and publish website]

Loading

The current270-cell matrix contains ten document/vector/queue/graph/stream scenarios. Intensive TimeSeries6/30-cell host dispatch, emitted wire, collection, aggregation and site joins remain pending under ADR-059. Image/model checks are not TimeSeries performance measurements. Source graph regressions and real exact-SHA GitHub jobs/artifacts qualify this repair; local source checks do not.

Isolated TimeSeries input source join

REQ-BC059/061/064 and AC-TH009-001..004 under ADR-059 now have additive strict private Host settings and one-selected-family Aspire composition source. The source receipt binds10 files and126 authored unit/model parameterizations; scoped development builds/formatter pass. Exact-source GitHub reports, dispatch/original lifecycle, physical copies/ACK, native6/30 and complete cohort/site remain pending. Existing source/provenance fixtures are input checks and never observed native evidence.

flowchart LR
    Selection[Closed family cell] --> Model[One native group and runner]
    Model --> Input[Strict private host settings]
    Input --> Pending[Original native lifecycle pending]
    Pending --> Evidence[Required actual Actions evidence]

Loading

The additive TimeSeries family has a strict embedded6/30-cell contract and a settled-run JSON writer under ADR059. The writer retains every planned slot and original observed ACK/count/failure; unstarted attempts carry no invented timing. Its workloadSucceeded field preserves the original DTO predicate and cannot certify native copies or publication. Source review/development evidence is in repair039; exact-source TUnit/native host/envelope/copy/coverage/30-cell/site joins remain required.

The 035 source-only repair receipt records reviewed HTTP deadline ownership, observed Kurrent fatal precedence, closed TimeSeries selection/physical-resource models and17 separate original TUnit report destinations. Development source checks pass in a scoped projection; full delivered-source/native/fault/cohort/coverage/site gates remain open.

Digest-backed Docker execution continuation

TASK-KLEVENT-001..003 (REQ/AC-BC-005, AC-PERF-004, AC-KLEVENT-001..003) repairs KeyLoad's missing public session event readback seen in exact b21 JSON. Two source files delegate to existing SDK reads, return null for absence and preserve strict cardinality. A fresh actual RF3 target proves seeded/new/conflict/ cancel/following flows; root owns the hook/docs/gates and bounded worker owns only those source/new helper files. Full schema3 nine-engine35+31 scope stays required.

REQ-BC-001/003/005/009/019/021 and AC-PERF-006/009 now map to AC-IMAGE-001..007 under the Accepted ADR-034 image stage. The sole normal/TimeSeries runner and all RF3 hosts consume real digest-backed images prepared once per GitHub runtime job, with actual manifest-byte/source label proof, owned output mount/user and native terminal exit0. Required internal AppHost config is KeyLoad:ContainerImages:Server and Benchmarks:ContainerImages:LoadGenerator; missing/invalid refs fail safely and cannot select a host-process fallback. Root owns shared contracts, AppHost/CI/docs and final source/evidence review; bounded image tooling/runner Dockerfile worker scope is frozen in acceptance and task graph. TUnit resource/metadata assertions and actual registry/Aspire execution prove the criteria in GitHub only. No public engine API/schema/ACK change, external registry publication, local runtime, tool installation, secret logging or test/bound weakening. Full nine-engine, native Single/Replicated and six-profile scope remains required and unqualified.

TASK-ISO-030H under ADR-056 preserves REQ-BC-001/003/009/050/055/056 and AC-IMAGE-006 while repairing the native image importer's HTTP lifetime. Supplemental acceptance is explicit:

Criterion	Pass/fail and evidence
AC-IMAGE-LIFE-001	One real referenced AbortController deadline owns and awaits the original operation through terminal body work, clears in finally, and rejects invalid bounds before submission; no detached race or global keepalive.
AC-IMAGE-LIFE-002	Readiness attempt and poll fit the remaining overall30s; manifest fetch/headers/bounded body/cancel share the existing30s owner. Every existing image/source/digest/status/output oracle remains. Source review plus actual native import is required.
AC-IMAGE-LIFE-003	New ImageHttpDeadlineTests real Node child/native controller tests assert abort settlement/exit0, prompt success/fault exit, original error identity, no late abort, and invalid-bound rejection. No HTTP double; lifecycle evidence stays distinct from HTTP proof.
AC-IMAGE-LIFE-004	Exact-SHA existing ImageBundleRealTests and native import/preflight retain original image bytes/config/source/manifests/outputs/owned cleanup. Failed/skipped/missing cells fail complete270/publication.

Root owns same-slice image-manifest and its internal helper, shared docs and integration; Luna owns only NEW ImageHttpDeadline* unit tests/program after the frozen contract. Public API/data/SQL/SDK/cache topology changes are N/A: this is an internal tooling lifetime repair. Original d45 MongoDB n2 import failed before tests; the exact pending inner HTTP await is unretained. New native proof is pending.

flowchart LR
    GitHub[Exact source runtime job] --> Images[Verified final OCI manifests]
    Images --> Host[Aspire RF3 and runner containers]
    Host --> Native[Real native terminal gate]
    Native --> Artifact[Raw reports source run and image identities]

Loading

The site candidate's observed analyzer dependency failure loop is traced under REQ/AC-BC-027 in site acceptance and its working plan. TASK014/015/016 preserve all diagnostics/tests/native thresholds while repairing genuine decimal display parsing, independent token-start location oracles, cohesive test classes and uncovered real diagnostic flows. The exact ownership, staged join and rollback contract is in ADR-033; no source-only repair qualifies site runtime, native coverage or the full database.

Status: in progress. Owner: lead benchmark integrator. Product scope and authoritative boundaries remain in the root policy and product specification. ADR-034 defines engine/deployment/evidence contracts.

ADR-021 owns PostgreSQL as the primary general-purpose baseline and the equivalent workload/guarantee contract; specialized engines remain separate workload arms. Neither decision establishes a performance winner without successful matched GitHub evidence.

Requirement	Type / priority	Behavior and rationale	Acceptance
REQ-BC-001	Correctness / P0	Use one deterministic dataset and exact caller-visible correctness oracle for every supported equivalent workload.	AC-BC-001; stable hash and verified documents/vectors/graph/events.
REQ-BC-002	Topology / P0	Run all servers and the load generator in Docker under Aspire, preserving real KeyLoad RF3 and SDK calls.	AC-BC-002; container model and runtime evidence.
REQ-BC-003	Replication / P0	Add native replicated engine groups with observed membership/copy/acknowledgement state.	AC-BC-003; real connected cluster proof before timing.
REQ-BC-004	Engine coverage / P0	Add MongoDB, OpenSearch and KurrentDB to the existing six engines using free community scope.	AC-BC-004; nine honest profiles/support matrix.
REQ-BC-005	Event streams / P0	Add expected-no-stream single-event append and bounded single-event read with identity/revision/cardinality/payload validation.	AC-BC-005; four supported engine implementations and readback.
REQ-BC-006	Provenance / P0	Use exclusively successful GitHub Actions JSON as public benchmark data and automatically derive graphs.	AC-BC-006; immutable evidence, successful-source gate and no hand-coded measured values.
REQ-BC-007	Product presentation / P0	Explain KeyLoad clearly in README and present qualified profiles/charts with raw data and guarantee differences.	AC-BC-007; generated SVGs, public links and real first-render proof.
REQ-BC-008	Performance / P1	Optimize measured bottlenecks without changing contracts and report matching before/after evidence.	AC-BC-008; comparable CI reports and focused correctness/security proof.
REQ-BC-009	Qualification / P0	Keep all required gates and unsupported readiness/durability boundaries honest.	AC-BC-009; exact delivered-source GitHub results and status docs.
REQ-BC-010	Resource bounds / P0	Stream JSON and raw CSV reports without creating a complete second copy of sample data as text; preserve every attempt and the report schema.	AC-MP-010; real-file roundtrip/quoting/cancellation regressions and CI resource evidence under ADR-035.
REQ-BC-022	Tooling / P0	Keep embedded microbenchmarks externally consumable by BenchmarkDotNet's generated child while all source quality rules and real fixture lifetime apply.	AC-EM-001..004; real metadata/store/TUnit process Dry proof under Accepted ADR-047, never comparative/RF3 evidence.
REQ-BC-023	Correctness and ownership / P0	Replace fake harness verification with genuine pinned Neo4j public runner/native response/data checks and limit cleanup to acknowledged acquired resources.	AC-GH-001..007 under ADR-049, acceptance and ordered graph; implementation and exact-SHA GitHub proof pending.
REQ-BC-026	Time-series comparison / P0	Compare persisted KeyLoad and Timescale paths over one deterministic UTC sample set, and exercise ManagedCode.TimeSeries as an in-memory aggregation library with separate guarantee metadata.	AC-BC-026 / AC-TSC-001..006 under ADR-050 and acceptance; source implementation and full Release build are complete, exact-SHA GitHub proof is pending.

Slice map

• Harness/adapters: benchmarks/KeyLoad.Comparisons/Features/BenchmarkComparisons/; existing flat files are migration debt tracked by ADR-032.
• Embedded scenario library: benchmarks/KeyLoad.BenchmarkScenarios/Features/BenchmarkComparisons/; existing KeyLoad.Benchmarks retains only its typed executable runner under ADR-047. New boundary implementation/qualification is pending; exact accepted contract is ../ADR/ADR-047-embedded-benchmark-host.md and its ordered working plan.
• Aspire engine resources: src/KeyLoad.AppHost/Features/BenchmarkComparisons/; shared Program.cs composition has one integration owner.
• Tests: tests/KeyLoad.ComparisonTests/Features/BenchmarkComparisons/ and tests/KeyLoad.UnitTests/Features/BenchmarkComparisons/; existing fixtures are shared.
• Evidence/chart tooling: scripts/Features/BenchmarkComparisons/.
• Site: site/Features/BenchmarkComparisons/; existing root assets are migration debt and updated only under serialized integration ownership.
• Docs: this spec, ADR-034, implementation/comparative-benchmarks.md and the README presentation.
• Time-series profile: root-owned Aspire resource in src/KeyLoad.AppHost/Features/BenchmarkComparisons/; profile/targets under benchmarks/KeyLoad.Comparisons/Features/BenchmarkComparisons/TimeSeries/; isolated composition in benchmarks/KeyLoad.ComparisonHost/Features/BenchmarkComparisons/; real tests under tests/KeyLoad.ComparisonTests/Features/BenchmarkComparisons/TimeSeries/; ADR-050.
• Product persisted schema/public API changes: N/A; engines use isolated per-run benchmark databases/collections/streams. Core optimizations require their own business-slice contract and serialized ownership.

Immutable harness collection and naming repair

REQ-BC-020 maps to AC-BCT-001..006 in ../ADR/ADR-044-benchmark-immutable-contracts.md, the explicit task graph in its working plan and ADR-044. This accepted stage owns the nine diagnosed mutable collection properties/cached oracles and solution-only CLR naming migration. Valid schema3 JSON and Single configuration spelling, corpus, scalar vector arithmetic and engine/wire semantics stay stable. Required malformed array rejection is intentional hardening. Real JSON/configuration/corpus/file fixtures are authored first; exact-source GitHub engine/full suites remain pending.

PostgreSQL schema lifecycle and constant SQL

REQ-BC-021 maps to AC-PG-001..006 in ../ADR/ADR-045-postgres-schema-ownership.md, task graph and Accepted ADR-045. A duplicate run-ID target must never delete another target's schema after failed initialization. Constant client commands bind native values; a transaction-local server context, namespace lock and durable private owner marker preserve exact workload tables and measured operations while guarding cleanup. First-authored real PostgreSQL helpers reuse the existing pinned Aspire container after measurement; no local qualification or extra RF3 startup. All real tests, exact-SHA evidence and the documented commit- response fault qualification gap remain pending. Product persistence/API N/A: only transient benchmark lifecycle metadata changes; no database product contract.

flowchart LR
    CI[GitHub Actions measured SHA] --> Aspire[Aspire private Docker network]
    Aspire --> KeyLoad[KeyLoad real RF3]
    Aspire --> Engines[Nine engine profiles]
    Aspire --> Runner[Docker load generator]
    Runner --> Oracle[Shared corpus and correctness oracle]
    Oracle --> JSON[Raw JSON and CSV]
    JSON --> Gate[Successful source and completeness gate]
    Gate --> History[Immutable GitHub result history]
    History --> Charts[Generated SVG charts]
    History --> Site[README and Pages]

Loading

Preliminary execution graph

The highest-capability architecture review is complete. The frozen constructors, schema3, 35/31 support counts, Kurrent26.1.2 free-cluster boundary and six-profile publication contract are in ADR-034. Root owns shared contracts, target registration, runner, central config, workflows, README and docs. Adapter workers own separate engine files; tooling worker owns separate chart/evidence modules. External-resource helper ownership must avoid the concurrent Orleans/AppHost migration.

Task	Requirements / acceptance	Write ownership	Dependencies / join evidence
TASK-BC-ARCH-001	All REQ-BC / AC-BC	None; highest-capability architecture review	Read current contracts/policies; returns exact engine topology and feasible bounded ownership.
TASK-BC-ADAPTERS-002	REQ-BC-001/004/005; AC-BC-001/004/005	New MongoTarget and KurrentTarget files under the harness feature	Root contract packet ready; native SDK semantics and official sources; reviewed code and GitHub checks required.
TASK-BC-OPENSEARCH-003	REQ-BC-001/003/004; AC-BC-001/003/004	New OpenSearchTarget files under the harness feature	Root contract packet ready; exact-vector oracle and actual shard/cluster state; reviewed GitHub proof.
TASK-BC-EVIDENCE-004	REQ-BC-006/007; AC-BC-006/007	New scripts/Features/BenchmarkComparisons chart/data modules	Schema/provenance contract ready; no measured constants; validates actual downloaded successful CI JSON; GitHub publication verifies output.
TASK-BC-REPLICAS-006	REQ-BC-001/003; AC-BC-001/003	Existing Qdrant/Rabbit/Redis target files and new matching native-receipt helpers in the harness slice	Lead freezes endpoints/client ownership and native receipts; root owns Aspire groups and registration; source review plus real GitHub profiles required.
TASK-BC-CURRENT-REPAIR-009	REQ-BC-001/003/004/005; AC-BC-001/003/004/005	Only Kurrent-prefixed harness feature files	REVIEW-008 primary-source findings: native settings parser, distinct observed local members and after-corpus probe/checkpoint cut; root reviews every repair and GitHub qualification remains required.
TASK-BC-PUBLISH-SYMBOLS-011	REQ-BC-006/009; AC-BC-006/009	Only the seven existing scripts/Features/BenchmarkComparisons modules	REVIEW-010 requires named boundary symbols for CLI flags, paths, encoding, statuses and schema fields; preserve all parsing, validation, immutable-history and chart behavior, then join full diff and actual successful-CI JSON byte evidence.
TASK-BC-REVIEW-010	All REQ-BC / AC-BC	None; highest-capability read-only reviewer	Starts on complete source packets; waits for Kurrent-009 and Symbols-011, inspects every final diff and reports source approval separately from pending GitHub integration.
TASK-BC-INTEGRATE-005	All REQ-BC / AC-BC	Root-owned shared surfaces only	Join every reviewed complete worker and concurrent required AppHost/API migration; run complete GitHub qualification, publication and live UI verification.
TASK-MP-008B	REQ-BC-010; AC-MP-010	ReportWriter.cs plus new reporting helpers and real-file tests under Features/BenchmarkComparisons/	ADR-035 accepted; no runner/contracts/adapters edits; lead reviews schema, all-attempt retention, quoting and cancellation; execution in GitHub only.

Least expensive capable coding tiers must be chosen per SDK/protocol risk; ambiguous replication/security decisions escalate to the high-capability planner. Workers cannot change shared/public contracts, add packages/config, install skills, run local tests/benchmarks, commit or push. Terminal states: complete, blocked, failed or cancelled. Partial or unverified results do not satisfy a join. Every acceptance maps to the same-numbered requirement, ADR-034, the task row above and planned GitHub evidence; no pending result is a verified feature.

Current source evidence

Joined review record records COMPLETE source packets for Redis-006, Kurrent-009 and Symbols-011, the highest-capability REVIEW-010, exact source manifests and the lead's integrated static/data checks. All 29 generated files match the original historical output and all three successful GitHub raw JSON files are preserved byte-for-byte. This closes the listed source findings only. Every required real-engine, Docker, delivered-source CI, publication, UI and performance acceptance gate remains explicit and pending in that record.

Actual caller composition repair (2026-10-02)

REQ-BC-001/002/003/005/009/019/021 map to AC-PERF-001–009 in performance acceptance and its ordered task graph. ADR-034's Accepted repair continuation freezes the three actual RF3 endpoint bindings, authenticated Rabbit management client ownership and PostgreSQL public event readback. This repairs located caller failures; it does not close the existing nine-engine, native replicated, Docker load-generator or six-profile contracts.

The exact f627 run37057708780 raw smoke files contain no KeyLoad measurements: setup fails KeyLoadThreeRealEndpointsRequired. Rabbit QueueCycle fails RabbitManagementClientRequired; PostgreSQL append readback reaches the default interface NotSupportedException. Separately, native test completion fails the Aspire terminal predicate. One failure is not evidence for the cause of another. The retained 48-sample TimeSeries arm has one attempt per timed operation and a different external topology; its raw timings are diagnostic, not a qualified throughput, tail-latency or winner claim.

Task / acceptance	Owning paths	Caller-visible verification
TASK-PERF-002 / AC-PERF-004	Harness PostgresComparisonSession; ComparisonTests PostgresStreamPublicRegression and existing PostgresSchemaPublicFlow join	Existing real Aspire PG target: seeded, absent, append, conflicting duplicate/no corruption, cancelled and healthy following event reads; no new SQL or doubles.
TASK-PERF-003 / AC-PERF-002/003	ComparisonHost settings/constants/owner and host-only validation; UnitTests real-child startup/bindings/cleanup cases	Exactly three distinct HTTP(S) peer origins with index0 matching primary; safe early invalid config; actual Basic-auth management client; all existing precedence/cleanup/secret checks retained.
TASK-PERF-004 / AC-PERF-002/003	AppHost Features/BenchmarkComparisons/BenchmarkCallerBindings and existing composition call	Actual node endpoints and broker management/user/password references; native three-copy and queue-member proofs remain mandatory.
TASK-PERF-005 / AC-PERF-001/009	Private GitHub native evidence; root-owned durable receipt/status	Exact c486 full relevant main baseline run37060131271, byte hashes/counts/failing cases; source-only builds cannot qualify tests.
TASK-PERF-006/007 / AC-PERF-005–007	Lead-owned native resource graph/registration, Docker load generator, bounded six-profile CI	Full accepted nine-engine graph and six profiles; no interim six-engine checkpoint qualifies completion.
TASK-PERF-008 / AC-PERF-008	Future owning Search/ResourceExecution/StorageRecovery/ClusterRouting contracts	Owner confirmed SIMD/.NET intrinsics first, Rust only after profiling; correct native ZoneTree APIs and bounded Orleans parallelism preserve atomic apply/read cuts/faults.

Implementation is staged: first-author real regressions, repair located bindings and public delegate, review each disjoint worker packet, build/format/governance, deliver all eligible current-main changes, then retain exact GitHub gates. No local tests/runtime/benchmarks, timeout increase, terminal-gate substitution, unpublished package, paid clustering or secret output. Numeric product coverage and all existing endurance/fault gates remain open until actual evidence exists. Product persisted schema/API N/A for this composition repair. Website delivery is owned separately under BC028 and retains its distinct source/evidence boundary.

flowchart LR
    Aspire[Actual Aspire resources] --> Bindings[Three peers and broker references]
    Bindings --> Host[Validated CLI composition]
    Host --> RF3[KeyLoad SDK and three-copy proof]
    Host --> Rabbit[Authenticated broker and queue proof]
    Host --> PG[Public PostgreSQL event readback]
    RF3 --> Report[Exact oracle and retained attempts]
    Rabbit --> Report
    PG --> Report
    Report --> Gate[Native completion and six-profile GitHub gate]

Loading

Product website and conceptual RF3 presentation

The owner requested a proper product design and Three.js. ADR-040, site acceptance and ordered plan own this bounded extension. The existing BC001–010 and all product/evidence criteria remain mandatory. Reader, keyboard/screen-reader user, constrained browser and evidence publisher are the actors; generated index is the entry, with independent graphics and report mounts. Backend/public API/persistence N/A: no database behavior changes.

Requirement	Measurable acceptance	Test or review mapping
REQ-BC-011: product-first responsive design	AC-BC-011: purpose/status/navigation visible at1440/768/390/320px in a single-row header; at ≤760px the same links open from a visible menu control and close after a selection without losing the section scroll. No page overflow, wrapped header or clipped controls; select text fades before its chevron key; labelled table scroll only	Qualified H: SiteBuildTests plus explicit real-browser desktop/mobile visual review
REQ-BC-012: bounded conceptual Three RF3	AC-BC-012: Three0.186.1 native backend recorded; one renderer/canvas, correct cancellation/disposal/resize/pause; DPR≤1.5,≤1M pixels,≤30 calls,≤5K triangles, no idle loop	Qualified H: vendor/build TUnit contracts; explicit real-browser graphics lifecycle/device evidence exception
REQ-BC-013: accessible independent content	AC-BC-013: semantic labels/focus/tabs and no-JS/reduced/coarse/unavailable-graphics retain text/poster/evidence access	Qualified H: SiteBuildTests plus required keyboard/no-JS/reduced-motion browser review
REQ-BC-014: unchanged historical arithmetic and controls	AC-BC-014: all6 scenarios/11 metrics/profile/repetition/log/whiskers/table/downloads match independent oracle over authentic successful-CI reports	Qualified H: SiteMeasurementTests/SiteMeasurementOracle using actual JS child process; real browser control review
REQ-BC-015: atomic validated report state	AC-BC-015: invalid/path/hash/rapid/error loads never mix data/provenance/downloads; abort and stale generation fencing	Qualified H: SiteEvidenceValidationTests over real production validators and controlled corrupt inputs; required rapid/error browser review
REQ-BC-016: canonical complete asset/build migration	AC-BC-016: feature-owned assets, exact verified vendor/raw bytes and clean isolated output; authored JS≤40KiB gzip/CSS≤20KiB gzip; vendor lazy and separately recorded	Qualified H: SiteBuildTests and source/static/vendor/size/raw-byte checks
REQ-BC-017: GitHub TUnit site qualification	AC-BC-017: independent centrally pinned TUnit/MTP/net10 suite, real bounded Node probe, preserved old edge intentions; only exact GitHub run qualifies	Qualified H: KeyLoad.SiteTests full suite in GitHub; explicit visual/lifecycle review exception; numeric coverage is recorded in the scoped website closure below
REQ-BC-018: honest source/evidence/delivery	AC-BC-018: site/measured revisions, successful run and hashes distinct; no historical→current/schema3/publication/CLI-model inference	Qualified H: SiteEvidenceValidationTests and strongest complete source/evidence review

Canonical site files, mounts, TUnit project, limits, task ownership and terminal joins are frozen in ADR-040 and the site plan before coding. Positive flow: read product→inspect actual report→download raw data. Negative/error: corrupt/unavailable report clears all dependent state; GPU failure stops decoration only. Edge: rapid profile changes cannot reintroduce stale data; reduced motion and narrow/no-JS views remain readable. No source/module name is a passing test or public publication result.

Site coverage and real-browser qualification extension

Requirement	Measurable acceptance	Test or review mapping
REQ-BC-024: measure mandatory authored-site coverage	AC-BC-024: exact-source GitHub execution reports at least80% authored JavaScript line coverage, at least70% available V8 block-branch coverage and at least90% line coverage for critical measurement/validation/build/provenance modules; missing required files fail. Raw ranges, source hashes, runtimes, conversion semantics and first baseline are retained.	TASK-SITE-COVERAGE-008: strict native V8 range/report regressions and after-session threshold gate; unchanged vendor/tests/assets excluded, no source ignore directives.
REQ-BC-025: exercise the generated site in a real CI browser	AC-BC-025: TUnit uses an already available real Chrome process and BCL CDP/WebSocket against real isolated emitted files/HTTP; controls, numeric correspondence, rapid/error/retry/provenance/download/focus and renderer lifecycle paths have assertions and precise coverage. Missing browser, incomplete source coverage or unexpected console failures fail; no installation, fake DOM/fetch, alternate runner or substituted measurements.	TASK-SITE-BROWSER-009: actual browser suite, retained version/coverage/source packet. Visual judgment and physically unexercised GPU loss retain the narrower manual evidence boundary.

The strongest coverage review found that manual design proof does not waive numeric root rules. These additions close that qualification gap; pending coverage is never a completed acceptance criterion. Their exact source inventory, converter semantics, task ownership and terminal joins are frozen in ADR-040 and the site working plan before write-capable workers start.

The actual SiteTests run36994330874 exposed a shared vendor-gzip portability failure before browser launch. REQ/AC-BC-016/017/024/025 retain their existing requirements; the strict byte-identity/current-runtime compression receipt and bounded persistent builder diagnostics are frozen in site acceptance, ADR-040 and TASK019/020 in the site plan. Real isolated-builder success/corruption regressions and specific intended negative errors map to those criteria. Every source/worker/strongest join and full GitHub Analyzer/native/site/Chrome/coverage gate is required; the failed baseline and missing hosted gzip length stay honest.

Fresh GitHub evidence and separate website deployment

REQ-BC-028 maps to AC-BC-028 in publication acceptance, ordered plan and ADR-040. The owner's subsequent request adds a separate post-test publication workflow; earlier design/DNS exclusions and all BC001–027 requirements remain recorded. Every number is read from actual comparison-job JSON. Select the highest successful comparison job by own-main-push run number and descending attempt history; unrelated database jobs or whole-workflow failure do not block website-only delivery. Exact comparison-step/artifact/ZIP proof and full current-main website qualification precede deployment. Separate sibling website/measured/control checkouts retain accurate revisions. Invalid selected evidence fails without fallback. Existing verified public data remains until a successful update. Main site/** changes and benchmark workflow completion trigger the separate action; workflow_run arrives after the enclosing workflow, not immediately at individual-job completion.

flowchart LR
    Producer[Actual CI comparison job] --> Artifact[Immutable JSON ZIP artifact]
    Control[Trusted separate Pages workflow] --> Select[Authenticated run job artifact proof]
    Artifact --> Select
    Select --> Inputs[Same verified ZIP to BCL qualification input]
    Inputs --> Qualify[Complete Analyzer site Chrome coverage gates]
    Qualify --> Build[Exact raw bytes and publication provenance]
    Build --> Fresh[Repeat current run attempt artifact proof]
    Fresh --> Deploy[Pages deploy only after success]

Loading

Canonical surfaces: scripts/Features/BenchmarkComparisons/github-evidence-*.mjs and github-evidence.mjs; SiteGitHub-prefixed tests/input helpers under tests/KeyLoad.SiteTests/Features/BenchmarkComparisons/; shared pages.yml; durable spec here. Backend/contracts/database persistence N/A: no runtime/data API changes. Actors, public receipt schema, positive/negative/edge/error flows, tests, exact worker ownership and terminal joins are frozen in acceptance/ADR/plan before writes. Pure controlled metadata/ZIP tests are not fake API transport evidence. Real workflow/provider/live evidence remains mandatory and is fulfilled for exact H by the scoped website closure below.

AC-BC-024/025 additionally retains actual CDP blank/no-script empty deltas with their raw hashes and zero line/branch contribution. Each session must independently map authored functions; empty-only/unmapped-only sessions and malformed/Node-empty receipts fail. TASK036/037's tests-first reader/protocol contract and actual G red baseline are in publication acceptance/plan and ADR-040. Every earlier source inventory, numeric threshold and complete real-browser assertion remains required.

Required analyzer dependency for the website candidate

REQ-BC-027 maps to AC-BC-027 in site acceptance, TASK-SITE-ANALYZER-COVERAGE-011 in the site plan, and ADR-033. The bounded website candidate must qualify its actual KeyLoad.Analyzers dependency using the complete real TUnit suite and native MTP18.11.2 Cobertura counts at the same candidate SHA. Freeze all30 source files and configuration; include all25 executable sources. Enforce module80% line/70% branch and each of12 critical diagnostic pipelines90% line coverage from integer counts. Missing, malformed, ambiguous, changed or empty evidence fails. Retain raw XML, exact source/runtime hashes, derived report and pure gate boundary/error regressions. This establishes only the first analyzer module baseline; broader AC-CQ-009, RF3 and broader repository no-decrease qualification stay pending. Mandatory website no-decrease coverage is retained in the scoped closure below.

Concurrent TimeSeries already owns stable REQ/AC-BC-026. The analyzer substage's initial draft collision was corrected to BC-027 before candidate delivery; TimeSeries requirements and their scope were preserved.

Preserving library and CLI prerequisite

REQ-BC-019 maps to AC-HOST-001..007 in host acceptance and Accepted ADR-043. The existing KeyLoad.Comparisons assembly/public API becomes a library; the new host owns sole CLI composition/lifecycle. Aspire retains its comparisons resource and existing configuration/topology. Public collections/enum/report/workload changes are outside this stage. Host tests use actual child processes; success/cleanup/cancellation qualification uses the real GitHub comparison suite. Private construction ownership requires explicit source review plus runtime evidence as stated in acceptance. Frontend/persistence/auth surfaces are N/A because this prerequisite changes only build/CLI ownership. Host task graph records exact disjoint scopes and lead-only join; existing nine-engine criteria remain mandatory.

Current website delivery consumes authenticated successful schema2 evidence and rejects unsupported versions without selecting older data after a successful comparison is chosen. Source inspection of main355's schema3 emitter is not successful producer or website qualification; failed/incomplete comparison artifacts cannot refresh the site. Its three profile folders and measurement-step names still match the publication transport contract. The current schema2 queue phase→PointRead control regression is TASK032 under REQ/AC-BC-014/025; complete real Chrome evidence is required, with every existing oracle and numeric gate.

Qualified website delivery, 2026-10-02

Website scope only: REQ/AC-BC-011–018/024/025/027/028. Exact H 6a82c86d0113270335368bbfbff0080ea1c1800a passed complete GitHub validation37011817610 and both separate automatic Pages publications: site-path push37013009381 and producer-completion workflow_run37014111869. Full118 analyzer and70 site tests pass without skips, together with all required native/JavaScript thresholds and source inventories. Canonical immutable evidence, design closure and publication closure map criteria, tasks, exact artifacts and actual provider/live proof; strongest TASK-SITE-REVIEW-007 documentation/evidence review is COMPLETE.

The live responsive product page includes the bounded independent Three.js scene, accessible charts/tables, workload/metric/repetition controls and exact JSON/CSV/Markdown downloads. Current published measurements are the authentic successful comparison36926803549 at9c570f8c33a7a9667507a8e1c0ca68860de3be45; failed current producer37013008931 supplies no new results. Website/control/measured revisions and run/job/artifact hashes remain distinct. Every performance number is derived from these raw reports; unsupported values stay unavailable.

Pages at https://www.keyload.cloud/ uses enforced HTTPS; live apex redirects301 to www. Both automatic triggers, current-main/evidence freshness, immutable Pages35-file output and all9 raw-report bytes were verified. Manual mobile/desktop/WebGPU review complements real Chrome qualification; physical GPU-loss qualification remains explicitly unexercised. The feature's global status, BC001–010/019–023/026, schema3/advanced profiles and database/endurance/readiness qualification remain unchanged.

REQ-BC-029: shared KeyLoad visual identity

AC-BC-029 owner revision, 2026-10-03 (design overhaul)

• Concept. The landing is an editorial product page that reads like a spec sheet, in the shared identity from ADR-053:

  • paper surfaces, oversized black display type and mono chapter eyebrows (01 — …);
  • graphite instrument panels;
  • Managed Code's iridescent marker as the only accent. There is no green or lime.

• Order. The page presents KeyLoad as the AI-native database for AI agents on .NET 10 and Orleans, keeps the early-development wording, and runs in this order:

  1. a hero with the live scene;
  2. a fact strip;
  3. why an AI-native database: the stitched stack against one database, and why we think one database is the future (AC-COMP-009);
  4. connected data in one request (DatabaseComposition);
  5. the anatomy of one agent call;
  6. a data-shape bento;
  7. an engine cross-section;
  8. why .NET, Orleans and ZoneTree (AC-COMP-009);
  9. the evidence instrument, with a perforated provenance receipt;
  10. a ledger of claims that are not made yet;
  11. the method;
  12. reproduce;
  13. the footer.

• Search metadata. The title is "KeyLoad — the AI-native database for AI agents". Description, Open Graph, Twitter and JSON-LD copy use the same AI-native positioning within the AC-SEO-002 contract; JSON-LD stays a two-node WebSite/SoftwareSourceCode graph without offers, ratings, reviews or actions.

• Agent call anatomy. A numbered timeline beside a graphite terminal shows the real MCP tool keyload_query_execute and the real .NET SDK atomic command. It describes the path, not live data.

• 3D scene motion contract (changed by owner direction).

  • The carousel of data-shape cards around the KeyLoad mark starts moving as soon as it is ready. It spins while visible and follows a fine pointer.
  • It pauses offscreen, on hidden pages and when the visitor pauses it. It never moves under prefers-reduced-motion.
  • AssertMotionPlaysThenSettles proves that render calls advance while playing and are identical after pause. Budgets are unchanged: 19 draw calls and 37 triangles.
  • The poster is a pre-rendered frame of the same scene on the stage gradient. The canvas is transparent, so the scene sits on the glass stage.

• Comparable engines only. The chart and table omit engines that do not implement the selected workload, and KeyLoad is highlighted. Their DOM rows, downloads and the engine count stay intact.

• Attribution. The footer reads "Developed by Managed Code" and links to https://www.managed-code.com/ with a normal followed link (no nofollow).

• AC-BC-029: the public site uses the shared light identity from ADR-053:

  • the same logo, tokens and sans display type, plus cards, buttons, tabs, bars and tables;
  • brand-palette engine colours and scene colours.

• site/Features/BenchmarkComparisons/brand.css and site/favicon.svg are byte-identical mirrors of the console's canonical brand.css and logo.svg. The test is SiteBrandParityTests.AC_VI_001_SiteAndConsoleShareByteIdenticalBrandSources in pages.yml.

• Every existing site hook, the size budgets, no-overflow widths, poster/scene lifecycle, zero console errors and the JS inventory remain unchanged. Copy keeps the early-development and no-winner wording.

• Subjective visual quality is a desktop/mobile screenshot review. It is not a numeric gate.

flowchart LR
    Canonical[Console brand.css and logo.svg] -->|mirror| SiteBrand[site brand.css and favicon.svg]
    SiteBrand --> Pages[Pages build BUILD.assets]
    Parity[SiteBrandParityTests] --> SiteBrand

Loading

Owner-directed isolated Linux performance matrix, 2026-10-03

The additive owner-requested favicon/search/social-preview presentation contract is SiteMetadata, REQ-SEO-001..006 / AC-SEO-001..007 under ADR053. It preserves every benchmark/evidence gate.

ADR-056 and the canonical acceptance/plan replace the three-OS and all-engine-on-one-runner producer for new qualification. Historical evidence retains its original topology/source/format; it is not qualification of the new matrix. Source and full GitHub/publication gates are pending.

Stable requirement	Acceptance	Owner and automated evidence
REQ-BC-050 Linux-only complete qualification	AC-ISO-001	TASK-ISO-007/010/012; workflow inventory + actual fullLinux CI
REQ-BC-051 one isolated agent per engine/node/scenario	AC-ISO-002	TASK-ISO-007/009/010; closed270-cell plan and native resource inventory
REQ-BC-052 actual native1/2/3 and honest unsupported topology	AC-ISO-003	TASK-ISO-005/009/010; independent membership/copies/ACK proof
REQ-BC-053 explicit benchmark fixed-voter safety	AC-ISO-004	TASK-ISO-005/010; negative configuration, SDK/MCP restart/quorum loss andRF3
REQ-BC-054 intensive shared CRUD correctness and raw failures	AC-ISO-005	TASK-ISO-008/009/010; deterministic plan/oracle + real native full body/cardinality/absence
REQ-BC-055 exact per-worker versioned JSON/provenance	AC-ISO-006	TASK-ISO-008/010; strict single-target/scenario report and actual job identity
REQ-BC-056 complete authenticated aggregation	AC-ISO-007	TASK-ISO-006/010; real Node/files/TUnit corruption checks + GitHub job/artifact manifest
REQ-BC-057 site metrics derived from isolated raw JSON	AC-ISO-008	TASK-ISO-011; full site validators/oracles/browser/native coverage
REQ-BC-058 publish only fresh complete successful cohort	AC-ISO-009	TASK-ISO-011/012; authenticated aggregate/worker/archive freshness + provider/live proof

Canonical slice map: BenchmarkComparisons in Comparisons library, ComparisonHost, AppHost/Features, UnitTests/ComparisonTests/SiteTests/Features, scripts/Features, site/Features and this durable feature doc; shared workflow/architecture remain root-owned infrastructure. Benchmark voter validation remains ClusterReplication with ADR-007. Other public SQL/MCP contracts areN/A changed because existing native operations/authorization remain. No duplicate feature behavior belongs in a layer.

flowchart LR
    Matrix[Engine node count scenario] --> Worker[One isolated Linux worker]
    Worker --> Proof[Native correctness and timing JSON]
    Proof --> Aggregate[All cells authenticated and complete]
    Aggregate --> Website[Generate metrics and qualify site]

Loading

The reviewed TASK-ISO-012P/012PB/012PJ source join is under Accepted ADR056, REQ-BC-056/057/058 and AC-ISO-007/008/009. Mandatory SiteCoverageGate prepares two authenticated immutable isolated ZIPs through BCL before source capture and rechecks the private original receipt bytes plus277 inputs after the full suite. SiteIsolatedCoverageInventoryTests exercises actual Node/V8 and rejects omitted current publisher sources. Every remaining production and critical source retains native80/70/90 gates. ADR-076 retires the obsolete historical collector and profile renderer; the source closure includes current producer tools and every actual builder/browser consumer, compared against trusted control source. The exact current original authority and website SHA must still match before Pages. Source readiness is not native coverage,270-cell, publication or live proof; actual partial failures are in the source qualification record under implementation.

Separate intensive TimeSeries family

ADR-059 and the canonical acceptance/plan specify30 additional isolated cells: KeyLoad/TimescaleDB × native1/2/3 nodes × Append/RawRangeRead/Latest/Aggregate/Windows, preceded by six private preflights. REQ-BC-059..064 map to AC-TSI-001..008 and TASK-ISO-TS005..011. Their exact corpus, direct timestamp/sequence ordering, native SQL ownership, real public negatives, ACK versus all-copy proof and bounded response validation are normative in the acceptance; the new family does not change the270-cell or historical48-sample wire. Native image feasibility and exact internal/SQL/provider contracts precede their implementation scopes. The frozen16-client/five-repetition/10000-operation profile records latency through complete decode and wall throughput including synchronous per-attempt validation. All30 raw results, authentic jobs/images and a separate complete family projection are required before new TimeSeries metrics publish.

Canonical technical ownership is BenchmarkComparisons/TimeSeries/Intensive in Comparisons, mirrored by new Unit/ComparisonTests files and new AppHost/host feature helpers; root owns existing selectors, workflows, JSON, site and docs. Product API/data migration isN/A: existing authorized SDK/MCP operations remain. Library-only memory aggregation isN/A in this native-node matrix; its historical semantic regressions remain. Requirements and all native/publication/coverage evidence are pending, with no qualified intensive TimeSeries result.

flowchart LR
    Seed[Identical4096 ordered sample corpus] --> Native[One engine and native topology per runner]
    Native --> Operation[One append read latest aggregate or windows scenario]
    Operation --> Verify[Bounded direct output and receipt validation]
    Verify --> Raw[Immutable30 cell family]
    Raw --> Site[Authenticated separate TimeSeries metrics]

Loading

Native baseline repair traceability is frozen in ADR056 TASK-ISO-026K/R/PG/M: REQ-BC-052/053/055 map to AC-ISO-002/003/004/005/006, original failing GitHub jobs and new genuine per-engine SDK/Redis-copy/PostgreSQL-slot/Mongo-auth regressions. Root joins these only in each selected engine's isolated topology job. The exact2f source/native receipt distinguishes passing source gates from11 failed native jobs and skipped270. Repairs are source implementation until same-SHA real native reruns pass; no site metrics or performance verdict follow from incomplete measurements. Current-main original baseline is separately retained in run37093992229 receipt. Exact2ec normal units have17 failures and RF3 has one queue-receive UnknownWriteOutcome; scalar and native comparisons are skipped. This does not erase the successful exact2f source gate or qualify the new repairs. Fixture identity/typed-assertion and queue fault work require their own causal repairs and final delivered-source GitHub proof.

REQ-BC-060..064 / AC-TSI-002..008: ADR059 TS008K-S accepts the additive real SDK adapter and pure actual-input tests before implementation. Native SDK/MCP/cancellation/membership/fault proof remains a separate6/30 family gate. The026/TS007B-D source checkpoint is main397a89c; actual run37097831105 fails17 normal fixture cases. Its RF363/recovery164/analyzer118 and pure runner36 pass; scalar/native comparisons are skipped. The original receipt is retained separately. A source checkpoint or development build does not establish measured performance.

TASK-ISO-030K maps REQ-BC-054/055 and AC-ISO-005/006 to AC-KC-030-001/002/003 in the root acceptance and ADR-056 accepted canonical teardown contract. The native1/2/3 full-volume fixture independently derives55378 real acknowledged streams, retains a foreign native event, applies120s/180s untimed production cleanup at concurrency16 and reads every actual tombstone. Root owns shared selector/evidence/workflow joins. Source/budget approval does not qualify the existing failed/deferred drain branch; genuine026KF process-boundary/fault proof remains pending. Exact delivered-source GitHub tests and authenticated complete workload artifacts are required.

TASK-ISO-031M traces REQ-BC-052/055 and AC-ISO-002/003/006/007 to AC-MR-031-001..005 in ADR056/root acceptance: exact same-image BSON integral fields, intended first writable primary and two fresh valid500ms rounds in existing120s; genuine20s lower-priority native election and observed automatic return under300s parent. Native1/2/3 plus persisted authentication regressions and immutable failed-source evidence remain mandatory. Four AppHost files plus NEW Mongo-prefixed tests have one implementation owner; root owns every selector/workflow/doc/evidence join. Source discovery or stable admission does not establish measured failover tolerance. The accepted ADR059 TS009S/TS007R packet adds a closed internal preflight/intensive TimeSeries selection, dedicated node-local composition context and real Timescale primary/physical-standby model for1/2/3. Matching selection and resource-model TUnit cases map AC-TSI-001/003/008; configuration and model assertions are source checks, not native role, copy or ACK proof. Root owns actual SDK/official MCP/Npgsql native6 host, separate raw family/protocol/workflow/collector/site joins before the30 measured cells can qualify. Old270 reports and productionRF3 stay under their existing contracts.

ADR056 TASK-ISO-032H/AC-HT-032-001..003 retains one referenced original HTTP lifetime timer until the original promise settles after abort. The real Node regression uses an unreferenced completion timer and preserves original result/reason plus process-exit assertions. Source code/build is separate from exact-SHA actual registry import and full27/270 provenance.

REQ-BC059/061/063/064 and AC-TSI001/003/006/007/008 now map staged AC-TB009-001/AC-TH009-001..004 to ADR059 TASK-ISO-TS009B/H-A/H-S/H-I. The TS009B source receipt records the actual incarnation parameter and canonical voter bindings, all three original model cases, development build/formatter and explicit native limitations. H-A owns one selected native composition and original family hash/cell identity; H-S owns closed private settings using shared actual source/image provenance. Neither dispatches the incomplete native lifecycle. Caller-visible positive, negative/edge/mixed/secret-safe input assertions are in the canonical acceptance; exact-source GitHub model/normal/scalar and later real6/30 checks are required. No new product public CLR, persistence, credential authority or defaultRF3 change. Frontend publication is N/A for these input stages because no new measured family is qualified; the complete native family site stage remains mandatory.

Shared comparison pipeline

The latest owner correction2026-10-03 uses exactly three workflows under ADR-064: ci.yml combines ordinary build/test/rule gates; benchmarks.yml (Benchmarks) owns every load/ comparison/TimeSeries check, complete JSON aggregation and the full website qualification/publication chain; release.yml builds and publishes real dated database delivery. REQ/AC-PIPE-001..004 and REL-001..003 are defined by ReleaseDelivery and its acceptance matrix. Website publication must authenticate the exact current benchmark run/attempt/source without historical fallback; recheck that tuple and current website source before deployment. Native isolated job/step/artifact identities, complete workloads and topology remain required; authentic historical legacy CI archives are not relabelled.

The accepted TS009W stage in ADR-059 retains all1280 real warmup attempt/ACK/sequence slots for REQ-BC060/061/062 and AC-TW009001..003; pure source tests and exact-source normal/scalar qualification remain separate from the still-undelivered30 native-cell warmup/copy oracle.

The additive native serialization diagnostics map REQ-IS-PERF-001..004 to AC-IS-PERF-001..004 under ADR060/ADR047. They are local development profiling; internal codec jobs and dispatch modes are removed from Benchmarks. Public figures require the complete native database cohort.

Lossless series-codec development controls

REQ-BC-CHUNK-001 maps to AC-CHUNK-006 in TimeSeries, ADR-079, TASK-CHUNK-MEASURE and TASK-CHUNK-JOIN. The shared microbenchmark feature owns public generated-consumer SampleChunk codec fixtures in KeyLoad.BenchmarkScenarios/Features/BenchmarkComparisons/; the existing typed KeyLoad.Benchmarks switcher remains their CLI. The compared code paths are current per-record native SampleRecord encoding/decoding and the bounded candidate chunk codec. Retain exact actual bytes/sample, source, corpus, settings, machine, cost and allocations. Fixed small batches are explicitly microbenchmark controls, never the required100k/1m/5m database datasets, RF3/durability evidence, public website comparisons or an acceleration claim. Actual canonical rewrite cost and correction recovery remain pending KL-078. UnitTests/Features/BenchmarkComparisons/SampleChunkBenchmarkConsumerTests.cs verifies all36 matched cases through the real external generated consumer; its Dry execution proves interoperability rather than performance. scripts/Features/BenchmarkComparisons/sample-chunk-development*.mjs binds the actual child execution, all36 original BDN cases and9 corpus manifests to unchanged source and Release dependency inventories. Its development receipt retains the original settings and explicitly rejects public or database-scale qualification. The codec development receipt retains the actual normal/scalar Aspire reports, two Dry consumers with copied negative cases and the ordinary36-cell matched codec control. Canonical storage, rewrite/recovery, representative database scale and Linux/RF3 gates remain open.

Failed-cell publication repair, 2026-10-04

The owner separates database measurement from browser/site qualification. A separate static-site build action may follow the generated JSON or run through CI; the latest clarification permits an end-of-Benchmarks static build. The selected implementation uses independent CI website jobs on own-main push/manual and completed Benchmarks events. Every build selects the newest completed own-main push/manual benchmark with a successful authenticated aggregate, across both producer event kinds. Invalid latest evidence fails without older fallback; a workflow_run trigger is authenticated separately from the selected producer. Website failures do not block metrics. Exactly CI, Benchmarks and Release remain. ADR-080 owns this workflow/trust-boundary change.

Owner direction explicitly supersedes the all-success restriction of REQ-BC-058, AC-ISO-007/009 and ADR-076 for benchmark publication. Every planned cell remains accounted for in the same authenticated run/attempt/source. Successful independent cells remain measured; terminal failed workloads have disposition: failed, reason: Benchmark failed; no measurement data is available., report: null, and the original failed job/step conclusions. No metric, winner or zero is inferred. KeyLoad engine repair and concurrent series-codec work are outside this task.

Requirement	Acceptance	Verification
REQ-BC-FAIL-001 independent workloads finish	AC-BC-FAIL-001 failed matrices do not skip aggregate; no fail-fast or arbitrary parallel cap	workflow source regressions and actual GitHub jobs
REQ-BC-FAIL-002 explicit unavailable cells	AC-BC-FAIL-002 workload failure produces a bounded null-report envelope before artifact upload; original job failure remains visible	real Node/files TUnit tests, actual failed-job artifact
REQ-BC-FAIL-003 authenticated partial results	AC-BC-FAIL-003 failed job accepted only with failed workload, successful result upload and matching failed envelope; missing/malformed/expired/mixed evidence rejected	producer/aggregate/site negative and positive TUnit regressions
REQ-BC-FAIL-004 honest site	AC-BC-FAIL-004 successful competitors retain values; failed cells have no numeric values and expose actual job link	independent numeric oracle, projection validation and real Chrome
REQ-BC-FAIL-005 repair shared preparation	AC-BC-FAIL-005 diagnose exact failed logs, repair benchmark setup/build invocation, retain native isolated topology and Aspire ownership	exact failed-source log, focused regression, delivered-source GitHub rerun
REQ-BC-FAIL-006 bounded registry readiness	AC-BC-FAIL-006 each native HTTP probe has at most2s within the unchanged30s total; only settled non-aborted HTTP200 succeeds; private no-follow diagnostics remain at most121 records/64KiB and evidence-write failures propagate. Actual fixture listeners use kernel-assigned loopback ports; optional numeric port accepts only integers1..65535 before HTTP/evidence, while both production callers keep the default5000	ImageRegistryReadinessTests:10 actual loopback HTTP/error/bounds cases, including rejected port inputs before I/O, plus genuine pinned Docker image export/import in GitHub
REQ-BC-FAIL-007 Kurrent writer starts after membership	AC-BC-FAIL-007 verify all native1/2/3-member views before constructing the SDK writer; retain native DNS seeds, TLS verification, leader preference, NoStream semantics, acknowledgements, replica-copy oracle and cleanup	IsolatedKurrentDiscoverySettingsTests:3 actual SDK/resource-model cases; genuine Aspire-owned StreamAppend preflights for1/2/3 nodes
REQ-BC-FAIL-008 explicit cancellation stops owned work	AC-BC-FAIL-008 workload, finalization and result upload use !cancelled() so ordinary failure still finalizes while cancellation stops execution/publication; always() cleanup retains bounded diagnostics and safely removes owned registries whose setup was cancelled	WorkflowBenchmarkFailureTests, unchanged canceled-job/producer rejection regressions and actual GitHub lifecycle
REQ-BC-FAIL-009 align native Redis transport	AC-BC-FAIL-009 each isolated Redis resource uses the documented native certificate opt-out for its existing authenticated RESP/TCP bootstrap; after actual Aspire startup, primary/replica endpoints retain scheme redis and native target port6379 with TLS disabled. Discovered mapped host ports remain dynamic; native client settings retain a password without ssl=true. Preserve health checks, wait dependencies, AOF always, native membership, direct-copy/cancellation checks and WAIT/WAITAOF acknowledgements	actual pinned resource-model regressions plus genuine Aspire-owned native Redis 1/2/3-node preflights; complete matrix and site qualification remain mandatory
REQ-BC-FAIL-010 preserve exact OpenSearch vector ordering	AC-BC-FAIL-010 the native query preserves the unchanged shared float32-input/double-accumulated cosine oracle and ordinal ID tie order. Use native scripted_metric map/combine/reduce over actual vector doc values; round parsed query values back to float32 before double arithmetic, retain at most TopK candidates per shard, merge only bounded shard TopK states, validate finite native double scores and retain projected content from the existing native source only for admitted candidates. Do not sort through float _score, weaken recall/tolerances, rerank/fetch in the client, duplicate stored vectors, use ANN or change the corpus/native topology/ACK contracts	independent deterministic precision witness for query d000000170 (documents2289/1272), positive/negative native query/response TUnit checks, and genuine Aspire-owned OpenSearch VectorExact1/2/3 jobs with every five10000-operation repetition passing the unchanged oracle; complete same-run aggregate/site/Pages still required
REQ-BC-FAIL-011 create the qualification evidence parent	AC-BC-FAIL-011 initialization creates the real workspace artifacts/site-evidence parent before any capture redirection, preserves existing directories/files, rejects file or symlink collisions at artifacts or site-evidence before output, and leaves isolated-capture absent for its exclusive producer. Never skip/rebind source/archive/270/277/test/coverage/browser/freshness gates	SiteQualificationStartupTests executes the actual initialization Bash block in fresh owned filesystems (new/existing parent, file/base-link/parent-link collisions), checks actual envelope redirection and unchanged targets, and runs as a mandatory Aspire-owned unit preparation gate; genuine new-source website qualification/Pages remains required
REQ-BC-FAIL-012 execute the actual native probe contract	AC-BC-FAIL-012 C# probe requests serialize the existing lowercase operation/values contract; median arithmetic and odd/even/empty assertions remain unchanged. Native error probes preserve string error codes and use the actual native error name for numeric DOMException codes; real AbortController cancellation must still assert AbortError	Existing SiteMeasurementArithmeticTests and SiteIsolatedHttpTests, original failed TRX and focused genuine-input development checks
REQ-BC-FAIL-013 complete disposable builder source	AC-BC-FAIL-013 SEO/vendor scratch repositories copy the exact canonical isolated-contract.json bytes from the actual source repository before invoking the actual builder. Vendor-only mutations use the accepted fixture's unchanged full original aggregate and270-worker inventory; they must not substitute an aggregate-only directory or duplicate the2GB raw inventory for each metadata mutation. Existing missing/link/corrupt-asset negatives must reach their intended checks before output; the valid vendor compression case must succeed. No fabricated contract, provider input, replacement builder or weaker assertion	Existing four SiteMetadataRejectionTests and three SiteVendorBuildTests, with actual source-byte copying and unchanged input authority
REQ-BC-FAIL-014 bound heavy qualification children	AC-BC-FAIL-014 expensive full-cohort builder/archive-verification children share bounded admission before their existing300s active execution deadline starts. Retain complete277 inputs/270 workers, stdout/stderr bounds, caller cancellation and original-byte verification. Hold ownership until actual child exit and readers settle; cancelled queued requests start no child. Select limits from original timings and bounded source-exact development measurements, then require every complete site test and native coverage gate in genuine Linux GitHub	Real-process admission/cancellation/failure regressions, original-input single-versus-parallel phase/CPU/RSS development receipts, complete unfiltered Aspire-owned GitHub site/TRX/Node+Chrome coverage and Pages
REQ-BC-FAIL-015 emit a confined browser catalog URL	AC-BC-FAIL-015 the actual built HTML uses data/isolated-catalog.json, which the existing strict confinedUrl accepts. Retain rejection of dot segments, traversal, foreign origins, credentials, redirects, mismatched hashes and unavailable evidence. Real Chrome must reach the existing Ready predicate and preserve every numeric/null, provenance, responsive and lifecycle assertion	Existing SiteIsolatedBuildTests, SiteIsolatedBrowserTests, SiteIsolatedFailedCellBrowserTests, SiteIsolatedStandaloneBrowserTests and SiteBrowserBehaviorTests; genuine Linux full site and coverage
REQ-BC-FAIL-016 validate the actual PNG-backed ICO structure	AC-BC-FAIL-016 distinguish the ICO file header type1 from each directory entry's color-count0 and reserved0. The canonical PNG-backed entries retain planes0, bits32, bounded16/32/48 frames and actual8-bit RGBA PNG payloads. Preserve exact console/site/output bytes, frame bounds, dimensions and missing/corrupt asset checks	Existing SiteMetadataTests.AC_SEO_001_IconsAreValidSharedBytesAndEmittedFromTheRealBuild; genuine Linux full site and coverage
REQ-BC-FAIL-017 database-only benchmark graph	AC-BC-FAIL-017 Benchmarks has only native preparation, database comparisons, authenticated result validation and JSON aggregation. No browser/site qualification/coverage/Pages dependency; a separate static build may follow ready JSON; every independent cell still finishes and failure still yields null data	WorkflowBenchmarkFailureTests and WorkflowThreePipelineLayoutTests; actual completed270 cells/aggregate JSON
REQ-BC-FAIL-018 independent automatic JSON consumer	AC-BC-FAIL-018 CI website jobs run on own-main push/manual and actual completed Benchmarks events, independently of ordinary CI/RF3 and producer workload failures. Consume only a successful aggregate and its exact original suite/provider archives. Chrome/site/build/deploy remain exclusively CI website work and cannot affect the producer	Updated workflow/source/producer-selection TUnit tests, genuine workflow_run CI and published JSON/browser join
REQ-BC-FAIL-019 authenticate the cross-workflow producer	AC-BC-FAIL-019 native Linux CI qualify/deploy context validates the real GitHub workflow_run payload: repository477801965, own-main head repository/branch, Benchmarks path/name, completed status, allowed producer event, immutable source, positive exact run/attempt and success/failure conclusion. Reject cancellation, fork, other workflow, changed tuple, missing/malformed/nonregular/oversized event file and historical publication overrides before capture. Authenticate that trigger separately from the latest selected producer and current CI executor/control/website; reselect latest and reauthenticate before deploy	Actual native context/selection rejection tests and CI AcPipe003; original archive receipts,270/277 identity and existing freshness coverage

| REQ-BC-FAIL-020 latest metrics for every website build | AC-BC-FAIL-020 select the highest run-number completed own-main Benchmarks run with success/failure conclusion across push/workflow_dispatch producers; pending, skipped and canceled runs have no finished comparison cohort and are not candidates. Require success/failure producer conclusion, successful aggregate and all original270/277 proofs. Reject missing/corrupt/latest failed aggregation without older fallback. Own-main CI push/manual and workflow_run all use this selection; changed latest tuple before deploy prevents stale refresh | Real producer-selection positive/negative TUnit cases, native CI capture and freshness/provider join |

| REQ-BC-FAIL-021 independent website queue | AC-BC-FAIL-021 own-main CI push/manual uses run-specific workflow concurrency so an unrelated older ordinary CI/RF3 run cannot queue the website consumer. PR retains ref cancellation; website qualify/deploy use distinct job-level serialization with cancel-in-progress=false. All tests, source/latest JSON and freshness gates remain unchanged | WorkflowLayoutCiSourceTests and actual CI job start while older ordinary source tests run |

TASK-FAIL-SEPARATE-001 (root) records policy/requirements/ADR before edits. TASK-FAIL-SEPARATE-002 (root) removes site generation and qualify/deploy from Benchmarks, adds independent CI own-main push/manual and workflow_run website jobs, and prevents ordinary CI suites from running again for that event. Keep native database matrices untouched. TASK-FAIL-SEPARATE-003 (producer worker) adapts only the site GitHub context, capture/aggregate-step contracts and their native SiteTests to the real CI event and strict latest completed producer selection. Original historical receipts retain their actual legacy aggregate-step inventory; new producers bind the five database-only aggregate steps. Never rewrite archives. TASK-FAIL-SEPARATE-004 (workflow worker) updates only workflow TUnit regressions. TASK-FAIL-SEPARATE-005 (root) joins exact-source build/format/governance/Aspire checks, scoped commit/push, genuine benchmark JSON and independent CI publication. Root owns action policies, dependency/source inventory and shared docs. No worker may alter engines, raw measurements, authorization, topology or numerical gates.

The independent source-stage development receipt binds the full0-warning/0-error build, format/governance and34/34 real Aspire/TUnit workflow/startup regressions to the actual source candidate. The six native context modules pass syntax checks; delivered-source native CI, full site coverage and publication remain pending. The genuine73 cohort independently completed all270 cells and authenticated JSON aggregation despite16 failed/null KeyLoad workloads; successful-job counts include explicitly unsupported cells and are not measurement counts.

ADR-080 owns the boundary change. The site follow-up is FAIL-SITE-PROBES (native worker, four test-helper files and the obsolete SiteVendorTokens.cs scratch-aggregate token), FAIL-SITE-ADMISSION (producer worker, exact heavy-child lifecycle helpers and regressions after the root contract), then FAIL-SITE-JOIN/DELIVERY (root). Native worker owns SiteNodeProbe.cs, SiteIsolatedNodeProgram.cs, SiteMetadataRejectionTests.cs and SiteVendorTestScope.cs. The producer first profiles the unchanged actual source/original input before admission changes. The original TRX contains25 builder cases and12 produce-rejection cases, plus five native inputs/verify-inputs cases matched by the heavy classification. These42 cases permit40 pending requests with two active children; cases with multiple invocations issue them sequentially. This is an inventory bound, not an observed concurrent queue. The approved admission policy is two active children, a FIFO queue of at most64 waiters and a20-minute cancellable admission deadline before the unchanged300-second active deadline. Local source6e/original AA270 profiling with native V8 coverage measured44.014s for one builder and53.448/53.444s for two; full277-input verification measured48.428s single and59.707/75.799s concurrent. Peak RSS stayed below483MB per child in these bounded macOS development sessions; this does not measure GitHub resource saturation or qualify website performance. Heavy classification includes actual builders, produce requests and inputs/verify-inputs probes; light probes and native network capture retain their existing paths and deadlines. An unsettled child or reader poisons admission and rejects pending requests, rather than releasing unsafe capacity. Producer owns SiteHeavyChildAdmission.cs, SiteHeavyChildLease.cs, SiteHeavyChildTokens.cs, SiteHeavyChildClassification.cs, SiteHeavyChildProcessFixture.cs, SiteHeavyChildAdmissionTests.cs and SiteHeavyChildCleanupTests.cs; its existing-file edits are limited to SiteIsolatedNodeProcess.cs, SiteIsolatedBuilderProcess.cs, SiteIsolatedGitHubNodeProcess.cs and SiteIsolatedGitHubNativeProcess.cs, all in the existing SiteTests feature. The real-process regressions cover FIFO/capacity, queued cancellation, queue-full/deadline rejection before start, native start failure, active cancellation, output bounds and unsafe ownership fail-closed. Root owns requirements, workflow/source closure, integration, full checks and delivery. An original source-exact verification copy with an explicitly recorded candidate test patch can provide filtered local development evidence. Never relabel the immutable receipt, current HEAD, CI/GitHub executor or local results. Filtering does not satisfy the full site/coverage qualification; production gates stay intact. The final16-file harness candidate passed all26 selected genuine-input Aspire case assertions (16 existing,10 new lifecycle cases), without skips or cancellation. The filtered run retained its actual exit10 and one synthetic after-session coverage failure; it does not qualify the full site or publication. A separate actual main73 candidate passed the full Release build, formatter and governance. Exact source, native-input, runner and TRX identities are recorded in the development receipt. Ordered task graph: FAIL-CONTRACT (root, complete) -> FAIL-SITE (site worker), FAIL-PRODUCER (tooling worker), FAIL-PREP (root/diagnostic worker) -> FAIL-JOIN (root review/build/format/governance/Aspire tests) -> FAIL-DELIVERY (scoped commit, push, genuine complete GitHub benchmark run and Pages receipt). FAIL-CANCEL is a root-owned follow-up before FAIL-JOIN: preserve ordinary-failure finalization, stop explicitly cancelled work, extend owned-registry cleanup and test all three matrix branches. Registry deadline regressions accept a final bounded aborted probe as well as a settled HTTP503; total/probe bounds and no-ready assertions remain unchanged. The C# bridge reports original Node stderr before exit status. All workers have disjoint write scopes; root owns contracts, workflows, receipts, inventories and docs. Local verification is development evidence; publication requires authentic GitHub artifacts and full existing site qualification. Interrupted jobs without authentic result artifacts remain a publication blocker rather than fabricated worker evidence. Baseline run37154664616 has failures in container preparation, missing result artifacts and KeyLoad workloads. Original Redis job111299652762 retained an HTTP503 response in comparison-preflight-qualification-redis-n1-point-read artifact11286445994; the retry policy accepted only403/429 and rejected503. Bounded transient GET retries and null-report finalization repair this shared setup path. Local macOS development evidence: full Release solution build passed with zero warnings/errors; Aspire-owned producer/finalizer tests 2/2, GitHub evidence tests 84/84 and workflow tests 23/23 passed. Original successful PostgreSQL job111299653748/artifact 11285823608 and RabbitMQ preflight job111299652838/artifact11285807871 are retained as compact immutable test-only fixtures with original hashes, source and provenance; 48 original measured-report positive/negative probes passed. These fixtures never substitute for current publication input. Final formatting is recorded separately; unrelated concurrent SampleChunk formatting is outside this repair. Delivered-source workload/site publication proof remains pending.

The first repair run37158699545/source65e8bf59 passed the clean Linux solution build, formatter and pinned Docker image preparation/roundtrip. Actual website cell keyload-n1-document-update job111309324270/artifact11286559225 failed its workload, uploaded successfully and passed unchanged production job/artifact/ZIP/ envelope/agreement/site validators with its original null report. This is genuine failed-cell runtime proof; complete270 aggregation and Pages remain pending.

That run exposed two native preparation defects. Kurrent n3 job111309323415 started a listening registry but its first HTTP request consumed the entire30s probe budget; the underlying transport cause was not logged. FAIL-PREP-REGISTRY uses2s attempts inside the same30s total and records bounded safe probe facts. Kurrent n2 job111309323419/artifact11286971852 began writer construction while native election was still PreLeader. Actual pinned SDK1.4.0 constructor/selector inspection proves eager discovery can choose a live follower at that stage; the exact cached endpoint was not logged. FAIL-PREP-KURRENT moves construction after awaited complete membership verification. These are benchmark fixture repairs; neither changes a database engine, URI, retry/ACK contract or topology. Root records this contract before integration, then runs the focused Aspire unit/comparison cases, full build/formatter/governance and an exact-source Linux rerun before claiming the two repairs or publication qualified.

FAIL-PREP-OPENSEARCH is a follow-up under ADR-080. Sourcefabff69193f41c784f0b36b85c1f34d82fa9903d, run37166698745, OpenSearch n1 VectorExact job111334162158 failed exactly two ExactRecallOrProjectionMismatch operations (2489/6585) in every repetition; both select corpus query d000000170. Native startup, membership, copy checks and cleanup succeeded. Original failed public artifact11291956733 has SHA256 1823f7fc876ad3b3a385056a8f75ab6bcd1de2cdb53bbef977a9e2d4313f5088 and report:null; original qualification11291373059 has SHA256 070fd453c92cf373666c3e993990f9e06bac07d111eac0795ebcdc03ed86d4d2 and retains the rejected raw result. Returned neighbor IDs were not captured in this original failure, so its precise native rank is not inferred. An independent offline mathematical witness shows the shared double oracle places document2289 (0.27605485016755293) before1272 (0.2760548285997265), while both collapse to 1.276054859161377 under float32 translated-score ordering. That witness is development proof, not a native query result or publication input. The follow-up must use the native exact bounded aggregation API and qualify actual returned IDs through the unchanged complete workload. The current failed-cell run continues independently; any repair requires its own authentic source/run/attempt qualification. Pinned OpenSearch3.6.0/k-NN3.6.0.0 registers vector doc-values access for scripted-metric contexts but not NumberSortScript; generic vector get(int) throws. Numeric _script sorting cannot use that vector accessor in this pinned version. The selected native map/combine/reduce path avoids that unsupported API and float _score ordering. Source/response helpers remain benchmark-owned; unchanged native VectorExact jobs must prove the strategy and resource bounds.

FAIL-SITE-STARTUP follows the successful original270 aggregate in run37166698745. Actual Check website job111348160002 failed before tests at02:55:05.5179465Z: its first shell redirection could not open artifacts/site-evidence/isolated-capture-envelope.json. QualifySite initialization exported paths but never created their parent. The selected website source was e97bf30af7ecff5aff467385832ed89a70fe6ee7; control/measured source remainedfabff69193f41c784f0b36b85c1f34d82fa9903d. Pages was skipped; the successful aggregate is not publication evidence. Root owns the minimal action initialization and its mandatory startup test gate. Producer worker owns only the two dedicated unit fixture/test files. Filesystem-only development tests use disposable workspace variables and no provider, run IDs, source rebinding or measurement inputs.

Native-stage local development checks: initial full Release build and formatter passed; Aspire registry10/10 and actual SDK/resource settings3/3 passed. The new mandatory preparation workflow regression passed. The full24-case workflow filter had23 passes and exposed its old22-comparison-filter inventory expectation; source now enumerates the added23rd filter explicitly. Genuine source6ec9233 run37161833119 passed clean Linux build, formatter and common pinned-image export/import. Original artifact11288810076 contains registry10/10 and SDK/resource settings3/3 with no failures/skips. Kurrent native1/2/3-node StreamAppend preflights all passed; their original artifacts11288133701/11288840411/11288083670 retain actual distinct native members and five10000-operation successful repetitions per cell. This4096-record profile proves preparation/preflight only; required scale/cohort/site gates remain.

CI37161833095 passed normal2863/2863 units but failed scalar status-deadline: the fixture incorrectly required every final probe to settle as HTTP503, even though its remaining budget can legitimately expire. Original Node stderr was not retained by the assertion ordering, so that particular assertion remains an inference until the focused rerun. The test accepts only settled503 or aborted timeout(null/503), retains all bounds and forbids ready facts. Production readiness is unchanged. The bridge checks stderr first. Thirteen prior-epoch recovery tests lacked the executable skipped after that scalar failure;18 RF3 product failures remain outside benchmark scope.

Run37158699545 also exposed workflow cancellation being delayed by always() workload steps. FAIL-CANCEL changes only nine matrix execution/finalization/upload conditions to !cancelled(); ordinary failure still runs the same failed-envelope path. All owned cleanup remains always() with cancelled partial image setup accepted through unchanged ownership validation. Source review passed. Current JS registry fixtures passed10/10 normal and10/10 scalar through the existing Aspire-owned TUnit bridge; the new C# follow-up runner was not built successfully. Full Release, scoped Unit build and formatter were blocked by concurrent KeyCodec/CRUD source/test diagnostics. New workflow C# tests and actual complete publication remain pending; none of that engine work is staged in this repair.

The clean88e2cec snapshot passed full Release build, formatter and governance. Its actual full unit report has2821/2867 passes,46 failures, no skips/cancellation: workflow cancellation3/3 and all selected workflow-layout groups passed. One registry case failed with original EADDRINUSE; concurrent local tests and an observed system ControlCenter listener share port5000. FAIL-PREP-REGISTRY therefore uses an OS-owned ephemeral port for each actual HTTP fixture. Only a numeric port on the fixed loopback URL is configurable in the readiness helper; preparation, bundle publication, Docker mapping, manifest fetch and cleanup keep their existing default5000. The evidence-bounds case verifies invalid numbers/types produce no requests or evidence directory. Other full-unit failures include temp-path link rejection and a native allocation assertion; those are not repaired here. Reruns use an owned canonical temporary directory and retain the original failed report. Clean-snapshot failed-envelope producer/finalizer2/2 passed through Aspire. The chunk development child failed because that archive has no Git metadata; its original stderr proves the failed git check-ignore, rather than a codec defect.

The subsequent benchmark-only port-isolation candidate (archive of88e2cec plus its scoped patch) passed full Release build with0warnings/errors, formatter and governance. Its freshly built actual Aspire/TUnit registry scenarios passed10/10 normal and10/10 scalar, no skips/cancellation/timeouts, using an owned canonical temporary directory. The production callers still omit the optional port. Source/module/fixture/runner and original report hashes are retained in the local development receipt and status. Complete delivered-source GitHub publication is pending; these development results do not qualify a full cohort or website.

The original source6ec Redis preflights now prove a separate preparation defect: jobs111319875391/111319875411 started native Redis8.4.0 on plain TCP, but the Aspire13.6 healthcheck attempted TLS and blocked app.StartAsync until the60-minute case deadline. Both failed jobs finalized and uploaded null-report availability envelopes, then completed owned cleanup successfully. No workload measurements started. FAIL-PREP-REDIS uses the documented resource-local native WithoutHttpsCertificate() API to preserve the existing authenticated RESP/TCP contract; it does not alter a connection string by hand or disable certificate validation. Its experimental compiler opt-in is limited to that API and direct annotation verification. Root integrates the bounded benchmark-only change after the native source review, runs focused regressions and full build/format checks, then verifies genuine new-source preflights and complete publication. No Redis measurement repair or full-cohort publication is claimed before that evidence.

The Redis transport candidate (archive of3aeb8cb plus its scoped benchmark patch) passed full Release build with0warnings/errors, formatter and governance. Its freshly built actual pinned resource-model regressions passed3/3 through the Aspire/TUnit comparison entry point, without skips/cancellation/timeouts. The experimental opt-in is confined to the resource API and actual annotation checks. These model results do not prove Docker startup or workload execution; new-source native1/2/3 preflights and complete aggregate/site/Pages acceptance were pending at that checkpoint.

Delivered fabff691 source run37166698745 subsequently passed all Redis native 1/2/3 preflights and12 main CRUD cells, including600000 measured successful operations. All240 competitor originals were verified;96 numeric cells contain 4800000 successful samples. The full270 aggregate job111347065594 passed and preserved its15 failed/null-report cells (12 KeyLoad,3 OpenSearch VectorExact). Website job111348160002 failed before capture/tests because its evidence parent did not exist; Pages was skipped. This proves complete aggregation with failed cells, but does not establish publication or the required large-dataset gates.

The exact OpenSearch and evidence-startup follow-up uses committed e97bf30 source plus20 scoped benchmark paths, excluding concurrent engine changes. Full Release build passed with0 warnings/errors; formatter and governance passed. Fresh Aspire-owned TUnit checks passed query10/10, response25/25, actual Bash/filesystem startup6/6 and workflow gates7/7, with no skips, cancellations or timeouts. These are local development results on macOS27.0.1/.NET10.0.12; a fresh genuine Linux270 cohort, native precision/edge cases and complete website/Pages acceptance remain mandatory before claiming this follow-up delivered.

flowchart LR
  Plan[Complete native cell plan] --> Jobs[Independent Aspire workloads]
  Jobs --> Success[Validated measurements]
  Jobs --> Failure[Failed job and null report]
  Success --> Aggregate[Authenticated complete cell accounting]
  Failure --> Aggregate
  Aggregate --> Site[Qualified website]

Loading

Database job groups, 2026-10-04

REQ-BC-GROUP-001 maps to AC-BC-GROUP-001: Benchmarks exposes nine separately named database job matrices. Each contains only its own target, three preflight checks and all 30 canonical node/scenario workloads. Every cell retains its own Linux runner. Database groups run concurrently with only plan/image dependencies; steps inside each cell run in order. No shared cross-database matrix or parallelism cap. Aggregate waits for every group and retains failed/null results.

REQ-BC-GROUP-002 maps to AC-BC-GROUP-002: readable database/node/scenario job names must agree with authenticated job discovery, finalization, aggregation and website receipt validation. Canonical plan schema, cell IDs, workload counts, artifacts, measurements and topology remain unchanged. Frozen historical names remain valid only as exact cell-name contracts in original authenticated receipts; no result bytes are rewritten. ADR-080 owns the implementation contract.

TASK-BC-GROUP-001: root owns workflow, planner/name validators and documentation; worker owns disjoint source-contract TUnit assertions. Update mapped WorkflowLayout/WorkflowBenchmarkFailure/NativeSerializationWorkflow/IsolatedPlanCli assertions before joining. Owner explicitly requests no tests or benchmarks for this Actions layout task: verify YAML, JavaScript syntax, matrix inventory and diff statically; do not dispatch workflows. Runtime/native qualification is unverified. Backend/API/storage/transport and database tests are outside this change.

TASK-IMAGE-LIFE-PROCESS-IDENTITY-RACE

Root freezes this narrow unit-process oracle correction on2026-10-05 after the original full Aspire unit75g report:3247/3248 passed, with an actual Win32Exception while reading StartTime after HasExited in AcImageLife003CancellationKillsAndReapsRealNodeProcessTree. REQ-BC-PROCESS-IDENTITY-001 / AC-BC-PROCESS-IDENTITY-001 refine AC-IMAGE-LIFE-003: the original PID and captured UTC start identity stay mandatory. CaptureIdentity remains strict. A later identity observation may treat a Win32 failure as exited only when a subsequent HasExited on that same actual Process confirms exit. An unconfirmed/live observation preserves the original failure; a failed confirmation retains both original errors and fatal-runtime priority. No broad Win32-to-success handling, PID-only replacement or synthetic process is allowed.

Use one cohesive new IsolatedAggregateNodeIdentityObservation helper for the existing support and reaper probes. Luna query_wave privately owns only those two probe joins and that new helper. Preserve all original cancellation, process-tree kill/reap, stdout/stderr joins, exact streams,5s cleanup and test deadlines. The cancellation aggregate from the failed original run stays evidence; do not remove it as presumed expected cancellation. Existing five real Node lifetime cases provide the runtime regression. Root reviews, integrates and runs their actual Aspire unit caller, then full normal/scalar gates as required. ADR-056 records the implementation contract. No benchmark runner, native database, website, public contract or persisted data is changed; rollback restores the two original probes and removes the helper. Local tests remain development evidence.