Skip to content

Investigate intermittent current-session lock exhaustion during concurrent starts #80

Closed

Description

@CBEPX

One macOS merge-CI attempt failed while five background tasks started concurrently. The launcher exhausted the lock retries for current-session.json.lock before the polling phase.

Evidence: CI attempt 1, integration test survives aggressive concurrent polling while many background tasks finish. The error was Job state lock remained busy. The state-lock implementation and this scenario were unchanged from v1.7.12.

The same scenario passed in PR CI, a focused local repeat, and merge-CI attempt 2. This does not prove that the contention problem is fixed. Keep this separate from the confirmed null-PID fixture correction in #79.

Next step: reproduce concurrent launcher contention under controlled load, identify which lock holder exceeds the retry budget, and add a deterministic regression. Preserve ownership, identity checks, bounded waits, and atomic state updates. Do not weaken the lock or add blind retries without evidence.

Activity

  1. CBEPX commented on Oct 14, 2026

    @CBEPX
    OwnerAuthor

    Agent work: stage finished

    Client: codex · Session: codex-cc76-release-20261014-1238 · Claim: 7d48bd5f-61de-459f-9018-ca5fc3e00264

    Mode: write · Started: 2026-10-14T12:39:22+03:00 · Updated: 2026-10-14T14:20:22+03:00

    Stage: Completed: PR79 merged at 95cc226; main CI and all required mutation reports passed; v1.8.1 published with anonymous asset checksum verification, pinned plugin installed, 8668 state files preserved, Fable and Sonnet native smoke passed. No owned jobs or child processes remain. Restart Codex to reload. PR78 is a separate dependency follow-up.

    Scope in this repository: package.json, package-lock.json, .codex-plugin/plugin.json, CHANGELOG.md, docs/superpowers/plans/2026-10-13-reliability-advisor.md, .superpowers/sdd/2026-10-13-reliability-advisor, scripts/lib/mcp-capabilities.mjs, tests/claude-cli.test.mjs, tests/peer-companion.test.mjs, tests/hooks.test.mjs, tests/integration/claude-companion.test.mjs, tests/advisor-controls.test.mjs, scripts/lib/state.mjs, tests/state.test.mjs, stryker.shard.config.mjs, tests/mutation-config.test.mjs

  2. added
    agent:busyActive agent-work claim; ownership is recorded in the coordination registry
    on Oct 14, 2026
  3. removed
    agent:busyActive agent-work claim; ownership is recorded in the coordination registry
    on Oct 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions