[gdluck]

Requested by A

What is the problem

The startup sweep sends *#4*0## and *#16*0*5## to every gateway whose profile advertises WHO 4 / WHO 16, but on most firmware these two requests are NACKed by the gateway itself and never reach the bus (own-firmware-oracle, published verdicts; confirmed live on the F461 and on a MyHomeServer1, and on a physical MH201 by the committed issue_466 sweep trace). Two consequences: the sweep waits for answers that never come, and on SCS-only gateways (no answer to *#4*Z## from a standalone thermostat) a thermostat stays unknown after a restart until it next reports by itself, which can take a long time (#185 on an MH201, #257 on an F461).

This PR leaves the two requests out on the models whose firmware drops them (startup sweep and myhome.sweep_bus alike), lets each climate zone on those SCS-only gateways ask for its temperature with *#4*Z*0##, and restores current_temperature from the last state with a current_temperature_restored flag until the zone reports again.

Refs #185, #257, #674. No open issue is closed by this change. Cold-start hydration of the subordinate zones through the central unit (#0) is tracked separately in #813; the trace-corpus schema, timestamp causality, issue_674 forensics and fixture privacy-gate follow-ups in #814.

The per-zone *#4*Z*0## request rests on the published thermo-ts10.tsv row (forwarded on all twelve emulated gateways) and on the live F461 (answered in 119 ms on a populated zone), so this PR does not wait on OpenWebNet-HA/own-firmware-oracle#60; that PR is cited only for the F422 interface gate (*#4*Z#4#II*0##), which fails closed where it is unverified.

---

What is the solution

1. gateway.py: three refusal properties, the MyHOME fix(gateway): close idle command sessions before the gateway drops them #381 pattern (the per-model knowledge belongs in OWNd's profiles):
   • general_zone_request_refused reads getattr(profile, "supports_general_heating_status", None) first and falls back to GENERAL_ZONE_REQUEST_REFUSED (F450, F454, F459, F460, F461, H4684, MH200, MH201, MH202, MyHomeServer1) when the flag is None or absent. The MH201 is not in the oracle: the entry is the physical issue_466 sweep trace (review point 1, below). No OWNd issue exists for the flag yet; a follow-up OWNd profiles PR should carry it.
   • sound_status_request_refused reads profile.supports_who(16) first and falls back to SOUND_STATUS_REQUEST_REFUSED (F450, F459, F460, F461, MH202, MyHomeServer1). Plain OWNd master still advertises WHO 16 on the F461 and MyHomeServer1 although every WHO 16 frame is refused there; fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 drops it from those two profiles, after which profile.supports_who(16) alone is enough.
   • interface_zone_request_refused reads getattr(profile, "supports_interface_zone_requests", None) first and falls back to INTERFACE_ZONE_REQUEST_REFUSED (F450, F453, F453AV, F454, F459, F460, F461, MH200N, MH202, MyHomeServer1: the ten images on which *#4*2#4#01*0## is nack, silent in feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60; H4684 and MH200 forward it). The flag belongs in OWNd as well.
2. _discovery_frames skips a request the gateway refuses (DEBUG log). Unknown and forwarding gateways (F453, F453AV, MH200N, F455, Generic) are unchanged. The WHO 18 meter probes are not touched. myhome.sweep_bus applies the same gate through handler._refused_by_gateway; force: true sends everything, as services.yaml and services.md say (with force: false the service sends no WHO 4 frame at all on the listed gateways and triggers no per-zone poll).
3. zone_temperature_poll_needed: only on SCS-only gateways (supports_zone_status False). The profile flag wins when it is stated: supports_general_heating_status True means the sweep answers and no per-zone poll is needed, False means it is; only an unverified flag falls back to ZONE_TEMPERATURE_POLL, an explicit table (F450, F454, F459, H4684, MH200, MH201, MH202). F450 is in the supports_zone_status exclusion tuple (its *#4*1## row is the F454's: the request is expanded into four raw SCS dimension queries and the command session ends with a NACK), and F460 is not in the table: it expands *#4*Z## into one request like the F461 ($00499013000 in trace-corpus.tsv) and answers it, so it keeps the zone status poll.
4. climate.py: on those gateways a zone polls on add with OWNHeatingCommand.get_temperature (*#4*Z*0##) instead of returning early: one bus query, answered in ~120 ms on a populated zone, instead of the four that *#4*Z## expands into (~1.3 s per zone when a central unit answers them all, a NACK after 1.5-2 s when a standalone thermostat answers only some). F455 and other unknown SCS-only gateways still send nothing (the F455 multiple timeouts #674 test is kept). A zone behind an F422 (WHERE Z#4#II) is not polled where interface_zone_request_refused is True (F450, F454, F459, MH202 among the polling models); the H4684, MH200 and MH201 poll *#4*Z#4#II*0##. _polls_temperature is a pure boolean property (built from _temperature_poll_wanted and _interface_poll_refused); async_update writes the one-shot DEBUG line when it withholds the poll, also when a fan or humidity request keeps the update going. No _poll_answered callback is attached to the dimension 0 request on purpose: the *#4*1*0## NACK on 11 of 12 oracle images is the bus TX timeout of an unpopulated zone (the QEMU images have no thermostat wired; the live F461 NACKs its unpopulated zone 1 the same way after 2.07 s), not a verdict on a populated one, and the result stays uncounted until a command-session capture of *#4*Z*0## on an SCS-only gateway exists. current_temperature is restored from the last state and flagged current_temperature_restored: true until the first *#4*Z*0*T##; a malformed restored value is ignored.
5. discovery.py: a bus-discovered zone on such a gateway keeps its poll on add unless the revealing frame carried its temperature.
6. HL4684 and the OWNd aliases: _model_key maps the SSDP model string HL4684 (manifest.json; not an OWNd alias yet) to H4684, and supports_zone_status judges the configured string and the model OWNd resolves it to (get_gateway_profile(model).model_name), so HL4684, L4684, NT4684, 003598 (F454) and 03535 (MH202) are treated like the model they stand for instead of polling *#4*Z##.
7. Tests whose default sweep model was an MH201 drop *#4*0## from their expectations (test_blank_start.py, test_multi_gateway.py, test_multi_gateway_regressions.py; the delegated-WHO 4 follower case moves to an MH200N, whose firmware forwards the request); tests/test_issue_304_310_batch.py: the MH200 sweep expectation drops *#4*0##; tests/test_multi_gateway.py: the MH202 secondary's sweep_bus leaves out *#4*0## and *#16*0*5## (43 status requests instead of 45).
8. Docs: docs/configuration/runtime_behaviour.md (startup table and Day 0 matrix), gateways.md, climate.md, services.md, and hardware-compatibility.md, whose "Gateway Startup Discovery Capabilities" table now has two WHO columns: the profile's advertised WHOs (validated against the OWNd profile and the oracle, as before) and the requests the sweep actually sends (derived from a per-row sweep_omits list in scripts/update_gateway_profiles.py, which a test compares with _discovery_frames() for every row). MH200 and MH200N are separate rows, the MH201 row names its *#4*0## refusal and per-zone *#4*<WHERE>*0## poll, the H4684 / HL4684 has a row. Plus the b16 release-notes delta.

Companion PR: OpenWebNet-HA/OWNd#142 (draft): drops WHO 16 from the F461 and MyHomeServer1 profiles (the WHO 16 half of this PR becomes profile.supports_who(16) alone once it is released). The MyHOME companion of OpenWebNet-HA/OWNd#142 is #808 (draft): the discovery rows are derived from profile.supports_who, so it is green against the pinned OWNd and against OpenWebNet-HA/OWNd#142.

---

State Diagram

stateDiagram-v2
    direction TB

    state "1. Startup sweep and myhome.sweep_bus" as Sweep {
        [*] --> Request : *#4*0## or *#16*0*5##
        Request --> Sent : profile advertises the WHO and the firmware forwards it, before - always
        Request --> Sent : sweep_bus with force true
        Request --> Skipped : *#4*0## with supports_general_heating_status False, or None and model in GENERAL_ZONE_REQUEST_REFUSED, DEBUG
        Request --> Skipped : *#16*0*5## with no WHO 16 in the profile, or model in SOUND_STATUS_REQUEST_REFUSED, DEBUG
    }

    state "2. SCS-only zone without *#4*Z##, before - unknown until it reports" as Zone {
        [*] --> Restored : last state had current_temperature, current_temperature_restored true
        [*] --> Unknown : nothing restored
        Restored --> Polling : zone_temperature_poll_needed, flag False, or None and model in ZONE_TEMPERATURE_POLL
        Unknown --> Polling : zone_temperature_poll_needed
        Restored --> Waiting : zone behind an F422 and interface_zone_request_refused, DEBUG once on update
        Unknown --> Waiting : same gate
        Polling --> Reported : *#4*Z*0## answered *#4*Z*0*T##, flag cleared
        Waiting --> Reported : the zone reports by itself
        Restored --> Reported : the zone reports by itself
    }

Loading

---

Hardware Verification

• F461 firmware 2.0.11, 2026-10-23 / 2026-10-24: *#4*0##, *#16*0*5## (and the other 17 WHO 16 forms), *#5*0## are refused by the gateway itself in under 10 ms; *#4*2*0## on the populated zone 2 is answered with *#4*2*0*0239## and an ACK in 119 ms, while *#4*1*0## on the unpopulated zone 1 ends with a NACK after 2.07 s (the bus TX timeout the oracle images show on every zone). With this PR the sweep sends *#1*0##, *#2*0## and the WHO 18 probes, no *#4*0## and no *#16*0*5##; right after a restart current_temperature shows the last value with current_temperature_restored: true, cleared at the first *#4*2*0*…##.
• MyHomeServer1 firmware 3.87.13 (the oracle emulates 2.82.06), 2026-10-24: *#4*0## NACK in 51 ms (nothing on the event session), the MyHomeServer1 row of the general-zone table confirmed on 3.87.13; *#16*0*5## and *#16*101*5## NACK in 3 ms; per-zone *#4*35## is answered (5 frames), so the MyHomeServer1 keeps its zone poll and needs no temperature poll; its zones 35..70 report *#4*Z*0*TTTT## by themselves.
• MH201 firmware 3.3.6 (WHO 13 code 200), committed issue_466 captures: the sweep trace sent *#4*0## (frame 11) and no zone frame followed (the only WHO 4 traffic afterwards is three *#4*0#0*17## broadcasts and zone 105's dimension 15), while the sibling trace of the same plant shows zones 2, 3 and 4 reporting (*#4*4*0*0241##, *#4*2*14*0100*3##, *#4*3*12*0090*3##). That is the MH201 entry of GENERAL_ZONE_REQUEST_REFUSED. The MH201 3.6.27 fresh-discovery log in the repository carries no WHO 4 traffic at all (the plant has no climate zone configured), so it neither confirms nor contradicts it.
• The per-zone temperature poll on an actual MH201 / F450 / F454 / F459 / H4684 / MH200 / MH202 was not captured: those models rest on the oracle rows, the Climate error #185 field report and the F461's answer to the same request. No F422-routed zone was captured either; the interface gate rests on feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60 alone.

Live retest (2026-10-24)

Consensus label: confirmed for this startup configuration (one F461, standalone zone 2; other models not tested). Four starts of a local Home Assistant with the combined branches. The HA debug log is the evidence for the requests that were not sent: at each start it logged Skipping `*#4*0##`: F461 firmware refuses it., Skipping WHO=5 discovery: not supported by F461 profile. and Skipping WHO=16 discovery: not supported by F461 profile. (the WHO 16 line comes from OpenWebNet-HA/OWNd#142's profile, installed in that HA), and no *#4*0##, *#16*0*5## or *#5*0## appears among its sent frames; the bus monitor history of the same starts shows the sent requests *#1*0##, *#2*0##, *#18*51*51##, *#18*51*113##, *#2*05##, *#4*2##, *#4*2*11##, *#4*2*60##, *#4*#0*20##. Zone 2 showed current_temperature: 24.0 from *#4*2*0*0240## and no current_temperature_restored flag (the zone had reported). The *#4*2*11## NACK at every poll is pre-existing upstream code (climate.py, fan-coil dimension), not touched here. The review follow-up commits change nothing on the F461 path (its sweep, zone poll and sweep_bus gating are covered by test_f461_sweep_in_full and test_sweep_bus_leaves_out_the_refused_requests_unless_forced); they were not retested live.

---

How is it tested

• tests/test_startup_requests_per_firmware.py (70 cases): the sweep per model for *#4*0## and *#16*0*5## (15 models, MH201 and HL4684 included), the F461 sweep in full, the hardware-compatibility.md sweep column against _discovery_frames() for every row, myhome.sweep_bus on an F461 with force False (*#1*0##, *#2*0##) and True (all five general requests), the two profile-flag overrides, the zone_temperature_poll_needed table (15 models), test_profile_flag_wins_over_the_zone_temperature_poll_table (×4), the HL4684 key and the aliased spellings L4684 / NT4684 / 003598 / 03535, MH201 and MH202 zones send *#4*3*0##, an F422-routed zone on F450 / F454 / F459 / MH202 sends nothing and is logged once at DEBUG by async_update (nothing by __init__ or the property; also when a humidity request keeps the update going) while H4684 / MH200 / MH201 send *#4*3#4#01*0##, the interface profile flag over the table both ways, an F455 zone sends nothing, a central unit never polls temperature, a bus-discovered zone keeps its poll until a temperature arrives, the restored temperature and its flag, a malformed restored value, a cancelled dimension 0 write never reaches _poll_answered.
• Captured frames (same module): the three archives below are loaded with the loaders the other trace tests use. f461_live_2026-10-23 run 1: *#4*2*0## answered with *#4*2*0*0239## and an ACK in 0.119 s, *#4*1*0## NACKed after 2.071 s, *#4*0## refused locally; an MH202 zone 2 sends exactly the captured *#4*2*0## and the captured reply clears its restored temperature. myhomeserver1_live_2026-10-24: *#4*0## refused in 0.051 s, *#4*35## answered with five frames, the 33 zones 35..70 of run 1 replayed through zones on a MyHomeServer1 handler (no temperature poll, *#4*35## kept). issue_466 MH201 sweep: frame 11 *#4*0## with no zone frame afterwards, the sibling trace's zones 2, 3 and 4; an MH201 handler leaves *#4*0## out and its zone 4 polls *#4*4*0##, the captured *#4*4*0*0241## gives 24.1. The synthetic zone 3 cases stay: they pin the request a zone sends, not a plant.
• tests/test_gateway_profiles.py: the regenerated discovery table rows (both WHO columns) and the sweep_omits validator (a frame of a WHO the profile does not advertise, or a frame other than *#4*0## / *#16*0*5##, is a violation). tests/test_services.py::test_handler_refuses_frame_only_on_an_explicit_true: the sweep_bus helper never hides a request on a handler without the gate or on a double returning a non-bool.
• Fail-before / pass-after: 34 of the original 40 cases fail on the unpatched base; of the cases added for the first review, 15 fail on the branch before it; of this round's changes, 15 fail on the previous branch head (the MH201 sweep row, the table-vs-sweep check, the four F422 log-on-update cases, the humidity log case, the F455 interface case, the MH201 capture replay, the two test_gateway_profiles.py cases, and the four MH201 sweep expectations in test_multi_gateway.py, test_blank_start.py and test_multi_gateway_regressions.py); all pass with the change. test_issue_304_310_batch.py::test_discovery_asks_an_mh200_for_its_amplifiers is updated for the intended change; test_issue_674_f455_climate_trace.py (including test_gateway_supports_zone_status_property) passes unchanged.
• Full suite against the PyPI-pinned OWNd behaviour (plain upstream OWNd master, f04d8aa, on PYTHONPATH): 4631 passed, 1 skipped, 100.0 % statement coverage of custom_components (16233 statements).
• ruff check ., scripts/typing_ratchet.py (0 strict errors), scripts/verify_ha_standards.py, scripts/sync_documentation.py --check, scripts/manage_translations.py check: all pass.

---

Where are the traces

Trace PR: #799 (merged, ad8431ff); the MH201 captures are the issue_466 fixtures already in the repository.

Subsystem	Capture	Plant	Frames verified by this PR
Startup requests refused by the gateway itself	tests/fixtures/traces/f461_live_2026-10-23/myhome_trace_F461_live_2026-10-23.json, run run1_oracle_replay_readonly	F461 2.0.11	*#4*0##, *#16*0*5## (all 18 WHO 16 forms), *#5*0## NACK in under 10 ms, nothing on the event session
Per-zone dimension 0 on a forwarding path	same archive, run run1_oracle_replay_readonly (plant_reads, rows 521-525)	F461 2.0.11, zone 2	*#4*2*0## → *#4*2*0*0239## + ACK in 0.119 s; *#4*1*0## (unpopulated zone 1) NACK after 2.071 s (rows 374-376)
Bus-visible replies of the patched integration's sweep	same archive, run run6_event_session_ha_fixed_integration (an event-session rx capture: it shows replies, not what HA did not send)	F461 2.0.11	*#1*0##, *#2*0##, *#4*2##, *#4*2*60##, *#4*#0*20## answered on the event session; the omission of *#4*0## / *#16*0*5## / *#5*0## is verified by the HA debug log of the live retest above, not by this run
The same two refusals on a second gateway family	tests/fixtures/traces/myhomeserver1_live_2026-10-24/myhome_trace_MyHomeServer1_live_2026-10-24.json, runs run1_identify, run2_reads_thermo_sound_alarm_lights and run3_reads_lights_addresses_humidity	MyHomeServer1 3.87.13	*#4*0## NACK 51 ms (run 3); *#16*0*5##, *#16*101*5## NACK 3 ms (run 2); *#4*35## answered with 5 frames (run 2); zones 35..70 *#4*Z*0*TTTT## (run 1)
*#4*0## on a physical MH201	tests/fixtures/traces/issue_466/myhome_sweep_MH201_all_2026-09-25T19-11-35.json (frame 11) against myhome_trace_MH201_all_2026-09-25T19-05-42.json (frames 0-10)	MH201 3.3.6	*#4*0## sent, no zone frame follows; zones 2, 3 and 4 report on the same plant
Per-zone temperature poll on MH201 / F450 / F454 / F459 / H4684 / MH200 / MH202, F422-routed zones	no physical trace: oracle rows (below) and the #185 field report (MH201)	—	—

tests/test_startup_requests_per_firmware.py loads the first, fourth and fifth archives and replays their rows (see "How is it tested").

---

What does the Firmware Oracle / Encyclopedia say

1. OpenWebNet Encyclopedia (functional/who-4-temperature-control/addressing.md, dimensions.md): WHERE 0 addresses the general probes ("all probes"); DIMENSION 0 "measured temperature" is read-only, four decimal digits 0000..0500. The status request *#4*0## is the general form of the zone status request; the encyclopedia does not say which gateway firmware forwards it.
2. Firmware Oracle (published at results/mcp_index.json 53f9514):
   • *#4*0## (results/<P>/<V>/oracle/full/trace-corpus.tsv): nack, silent on F450 020010, F454 020051, F459 020105, F460 020012, F461 020011, H4684 020054, MH200 020102, MH202 010024, MyHomeServer1 028206; nack, out (forwarded) on F453 010100, F453AV 030014, MH200N 010108. That row plus the MH201 capture is GENERAL_ZONE_REQUEST_REFUSED.
   • *#4*1## (trace-corpus.tsv): the SCS gateways expand it into four bus queries, dimension 0, 12, 14 and 13 ($0499013000, $0499013012, $049901300E, $049901300D), nack, out on F450, F453, F453AV, F454, F459, H4684, MH200, MH200N and ack, out on MH202; F460 and F461 send a single $00499013000 and MyHomeServer1 one payload. Against a 99-zone central unit every query is answered (F454 + #0, tests/fixtures/traces/issue_629/home-assistant_myhome_2026-10-05T11-16-10.935Z.cleaned.log lines 489-700: *#4*1##..*#4*15## complete in ~1.3 s each with all five frames); a standalone thermostat answers only some, hence the NACK after 1.5-2 s. The 10 s socket hang of F455 multiple timeouts #674 was an F455 (not in the oracle) on actuator-only channels (*#4*Z*19*0*0##). That row puts F450 in the supports_zone_status exclusion tuple and keeps F460 out of ZONE_TEMPERATURE_POLL.
   • *#4*1*0## (results/<P>/<V>/oracle/full/thermo-ts10.tsv, also in tests/golden/firmware_oracle.json, the WHO 4 probe of scripts/update_gateway_profiles.py): ack, out on MH202 010024; nack, out with the bus payload 93 01 14 00 / $004990001000 on the other eleven images (F450, F453, F453AV, F454, F459, F460, F461, H4684, MH200, MH200N, MyHomeServer1). The per-zone dimension 0 request this PR sends is forwarded everywhere; the NACK is the bus TX timeout of an image with no thermostat wired, identical to the live F461's unpopulated zone 1 (rows 374-376), while its populated zone 2 answers with an ACK in 119 ms. Until a command-session capture of *#4*Z*0## on an SCS-only gateway (F454 / MH200) exists, no _poll_answered callback is attached to it.
   • *#16*0*5## (results/<P>/<V>/oracle/full/sound-who16-source.tsv): nack, silent on F450, F459, F460, F461, MH202, MyHomeServer1 (77 of 77 WHO 16 frames silent on each); ack, out on F453, F453AV, F454, H4684, MH200, MH200N. That row is SOUND_STATUS_REQUEST_REFUSED.
   • Open, not published: feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60 (thermo-zone-requests suite) adds zone 2, #0 / #23#1 and #4#01 coverage: *#4*2*0## out on every image (consistent with the thermo-ts10.tsv row above), *#4*2#4#01*0## nack, silent on F450, F453, F453AV, F454, F459, F460, F461, MH200N, MH202 and MyHomeServer1 and forwarded on H4684 and MH200, *#4*2#4#01## a thermoregulation-daemon crash on the MH202. That is INTERFACE_ZONE_REQUEST_REFUSED; the gate fails closed (no poll) on the listed models and sends on the two that forward it and on the MH201, which is not in the oracle.

---

Review follow-up (review 5642254816)

1. F450 joins the supports_zone_status exclusion tuple, F460 leaves ZONE_TEMPERATURE_POLL (now explicit) and the SCS-only list in climate.md; runtime_behaviour.md line 15 says which models poll *#4*<WHERE>## and which ask *#4*<WHERE>*0##; ("F450", True) and ("F460", False) added to the zone temperature poll table -> ede6721
2. F422 interface-routed zones: interface_zone_request_refused (profile flag first, INTERFACE_ZONE_REQUEST_REFUSED fallback), _polls_temperature off for such zones on those models with one DEBUG line, the _poll_answered omission commented, climate.md qualified, tests -> 329ec0a
3. myhome.sweep_bus with force=False skips handler._refused_by_gateway(q) frames, force=True sends them; F461 test both ways, MH202 multi-gateway expectation 45 -> 43, services.yaml / services.md / release notes -> 06afedf
4. HL4684 keyed as H4684 in _model_key and listed in the supports_zone_status tuple; the OWNd-aliased spellings (L4684, NT4684, item numbers) follow their profile's model name in supports_zone_status (found while cross-checking the diff); tests -> 8dcd9bc
5. Day 0 matrix WHO 4 / WHO 16 rows per gateway, GATEWAY_DISCOVERY_METADATA notes for MH202 (no more "Sound (*#16*0*5##)"), F454, F461, MyHomeServer1, MH200, hardware-compatibility.md regenerated -> a0ff59a
6. PR body: thermo-ts10.tsv row *#4*1*0## cited directly, feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60 kept for zone 2 / #0 / #4#01, the draft blocker sentence replaced, the state diagram updated for the interface gate and sweep_bus.

(The shas are those after the rebase on the current v2-phase1-architecture.)

Review follow-up (review 5642844379)

1. "MH201" added to GENERAL_ZONE_REQUEST_REFUSED on the issue_466 evidence (sweep frame 11 *#4*0## with no zone frame, sibling trace zones 2, 3, 4); runtime_behaviour.md lines 15 / 37, climate.md, the b16 release note and ("MH201", False, True) in test_sweep_skips_requests_the_gateway_refuses; the tests whose default sweep model is an MH201 drop *#4*0## (the delegated-WHO 4 follower case moves to an MH200N) -> e4d8adc
2. climate.py comments corrected: SCS gateways expand *#4*Z## into the four dimension queries (F454 + #0 answers in ~1.3 s, issue_629; the F455 multiple timeouts #674 10 s hang was an F455 on actuator-only channels); the *#4*1*0## NACK is the unpopulated-zone bus TX timeout (live F461 zone 1 after 2.07 s) while the populated zone 2 answers in 119 ms; _poll_answered stays off "until a command-session capture of *#4*Z*0## on an SCS-only gateway exists"; *#4*Z*0## kept; feat(climate): hydrate cold subordinate zones via one-shot *#4*Z## once live #0 central unit is confirmed #813 named as the cold-start #0 track; this body corrected the same way -> 3c2f5c9
3. _polls_temperature is a pure boolean property; the one-shot DEBUG line and _interface_poll_skip_logged moved into async_update, written when the poll is withheld (also when a fan / humidity request keeps the update going); tests assert nothing is logged by construction or by the property and once on update -> a4abfe6
4. hardware-compatibility.md: the column is split rather than trimmed, because calibrate_discovery_with_ownd recomputes queried_whos from the OWNd profile and the oracle validator requires the two to match; column 2 stays the profile's WHOs, column 3 is what the sweep sends (per-row sweep_omits, validated, and checked against _discovery_frames() by a test); MH200 and MH200N split, the MH201 note (*#4*0## refusal + per-zone *#4*<WHERE>*0##) and the H4684 / HL4684 row added; regenerated with sync_documentation.py --update -> 2af4e8b
5. services.md: myhome.sweep_bus with force: false sends zero WHO 4 frames on F450, F454, F459, F460, F461, H4684, MH200, MH201, MH202 and MyHomeServer1 and triggers no per-zone poll; force: true sends them and the firmware drops them -> da8c29b
6. tests/test_startup_requests_per_firmware.py loads f461_live_2026-10-23 (zone 2 *#4*2*0*0239##, zone 1 NACK), myhomeserver1_live_2026-10-24 (zones 35..70) and the issue_466 MH201 sweep (frame 11 *#4*0## with no reply) with the loaders the other trace tests use and replays the captured frames; the synthetic zone 3 cases stay and say so; the "Where are the traces" row of run6 credits it with the bus-visible replies only and the tx omissions are attributed to the HA debug log of the live retest -> d3be161 + this body

---

Checklist

• PR Description: Follows the structured sections above (What is the problem, What is the solution, State Diagram, How is it tested, Where are the traces, What does the Firmware Oracle / Encyclopedia say).
• Trace-Backed (if protocol/device change): Behavior is verified against real OpenWebNet bus traces or gateway firmware fixtures rather than guessed. (The two skips: live F461, MyHomeServer1 and the physical MH201 sweep plus published rows; the *#4*Z*0## fallback: the published thermo-ts10.tsv row and the live F461; the F422 interface gate: feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60, unpublished, failing closed.)
• V2 Architecture: New or modified entities implement handle_event() and do not manually subscribe via async_dispatcher_connect.
• Tests & Linters: pytest, ruff, and scripts/verify_ha_standards.py pass.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[GreenGrassBlueOcean]

@gdluck please rebase and solve conflcts

[gdluck]

Rebased on the current base (head f1b1d9e, mergeable): climate.py keeps both the merged #791 setpoint-in-force logic and this PR's zone temperature poll / restored flag, the release-notes line sits below the merged #791 entry, and the Known Limitations row from #791 came in untouched. Tests of both PRs run together on this branch (64 passed) and the full suite is 4538 passed at 100 % coverage; ruff, typing ratchet, HA standards and docs sync are clean. The other open drafts of the batch were rebased the same way.

[GreenGrassBlueOcean]

Solid protocol direction — verified GENERAL_ZONE_REQUEST_REFUSED and SOUND_STATUS_REQUEST_REFUSED 1:1 against own-firmware-oracle (trace-corpus.tsv, sound-who16-source.tsv, thermo-ts10.tsv), the live F461 (2.0.11) and MyHomeServer1 (3.87.13) captures, and OWNd==2.0.0b11.

Requesting changes for a few concrete issues before undrafting/merging (inline comments attached, plus two cross-file items outside the diff hunks below):

1. F450 / F460 mismatch (gateway.py, climate.md, test_startup_requests_per_firmware.py): Neither "F450" nor "F460" is in supports_zone_status's exclusion list, so zone_temperature_poll_needed returns False for both — causing legacy SCS gateway F450 (020010) to send *#4*<WHERE>## (#674 timeout hazard) instead of *#4*<WHERE>*0##, while climate.md misclassifies F461-sibling F460 (020012) as SCS-only.
2. F422 interface-routed zones (climate.py): *#4*2#4#01*0## is nack, silent (dropped locally by firmware) on F450, F454, F459, and MH202 in OpenWebNet-HA/own-firmware-oracle#60.
3. myhome.sweep_bus (custom_components/myhome/services.py:241-270): When force=False, sweep_bus builds its own general_queries list and gates only on _handler_supports_who(handler, who) without checking handler._refused_by_gateway(q). As a result, myhome.sweep_bus (force=False) still sends refused *#4*0## and *#16*0*5## frames and waits BUS_QUIET_PERIOD per NACKed frame in send_paced(). Please filter general by not handler._refused_by_gateway(q) when not force (or reuse handler._discovery_frames()) and add a test for force=False vs force=True.
4. Published oracle row for *#4*Z*0## (Draft blocker in PR description): The PR body states there is no published *#4*Z*0## row yet, blocking on own-firmware-oracle#60. Note that *#4*1*0## is already published across all 12 emulated gateways in results/<P>/<V>/oracle/full/thermo-ts10.tsv (and already in tests/golden/firmware_oracle.json, checked by scripts/update_gateway_profiles.py:502): ack, out (93 01 14 00) on MH202 and nack, out (93 01 14 00 / $004990001000\r / internal bus payload) on the other 11 models. You can cite thermo-ts10.tsv directly alongside #60 (which adds zone 2, #0/#23#1, and #4#01 coverage), so this PR doesn't need to wait on #60 once the items here are addressed.

[gdluck]

Thanks for the thorough pass; every item is on the branch (head 5985b69, rebased on the current base, one commit per point):

1. F450 / F460 (768f29b): "F450" is in the supports_zone_status exclusion tuple, so it polls *#4*Z*0## (its *#4*1## row equals the F454's: four raw SCS telegrams and a NACK). F460 keeps supports_zone_status True and leaves ZONE_TEMPERATURE_POLL, which is now an explicit table of the SCS-only models (F450, F454, F459, H4684, MH200, MH201, MH202), and the SCS-only list in climate.md; runtime_behaviour.md line 15 reads as you suggested. ("F450", True) and ("F460", False) are in the poll-table parametrization.
2. F422 interface-routed zones (1c5c651): new handler property interface_zone_request_refused, read from the profile first (getattr(profile, "supports_interface_zone_requests", None); that flag belongs in OWNd) with a MyHOME fallback table of the images where feat(oracle): add thermo-zone-requests suite (per-zone polls, actuators, humidity, interface form) own-firmware-oracle#60 has *#4*2#4#01*0## nack / silent. _polls_temperature is off for a zone with an interface on those gateways and logs once at DEBUG; H4684, MH200 and MH201 poll *#4*Z#4#01*0## as before. The async_update comment and climate.md are qualified, and the _poll_answered omission is now commented with the thermo-ts10.tsv reason (*#4*Z*0## ends in NACK on 11 of 12 even when forwarded, which cancels written).
3. myhome.sweep_bus (c9e1c3e): the general requests go through handler._refused_by_gateway(q) when force is False and are all sent when force is True, on the primary and follower lists alike. Tested on an F461 both ways (force=False → *#1*0##, *#2*0## only; force=True → all five); the MH202 expectation in test_multi_gateway.py goes 45 → 43 status requests; services.yaml / services.md describe force.
4. HL4684 (fec46c5): _model_key keys HL4684 as H4684 and "HL4684" is in the supports_zone_status tuple. While checking this, the spellings OWNd aliases (L4684, NT4684, item numbers such as 003598) had the same gap in supports_zone_status; it now also judges get_gateway_profile(model).model_name, with tests for each.
5. Docs (5985b69): the Day 0 matrix WHO 4 / WHO 16 rows are per gateway with the first-bus-frame note; GATEWAY_DISCOVERY_METADATA no longer claims "Sound (*#16*0*5##)" for the MH202 (77 / 77 WHO 16 inputs silent in sound-who16-source.tsv) and notes the dropped requests on F454, F461, MyHomeServer1 and MH200; hardware-compatibility.md regenerated.
6. Oracle citation: the body cites the published thermo-ts10.tsv row *#4*1*0## directly (ack / out on MH202; nack / out with 93 01 14 00 / $004990001000 on the other 11) and keeps unable to see Automation (cover) events #60 only for zone 2, #0 and #4#01, so the PR no longer waits on it. I will un-draft once CI on this head is green.

Full suite 4584 passed / 1 skipped at 100 % coverage; ruff, typing ratchet, HA standards, docs sync and translations clean; tests/test_startup_requests_per_firmware.py 60 cases (15 of the 20 new ones fail on the previous head).

[gdluck]

All checks green on 5985b69; marking ready for review.

[GreenGrassBlueOcean]

Thanks for working through the items from review 5642254816 (F450 / F460 split, HL4684 + OWNd aliases, F422 Z#4#II guard, sweep_bus(force=...), current_temperature_restored).

Cross-checking 5985b696 against both the existing physical traces in tests/fixtures/ and the newly ingested live captures (#799) leaves six concrete items to resolve before merge (inline below):

1. MH201 double-queries *#4*0## + *#4*Z*0## and overlooks two committed physical MH201 traces (gateway.py:158-171, runtime_behaviour.md:15/37, 2.0.0b16.md).
2. Keep *#4*Z*0## as the SCS-only startup baseline here, but fix the inaccurate protocol comments & PR body claims (climate.py:576-580, climate.py:597-604) that conflate #674's F455 10 s socket timeout and the busless QEMU oracle's NACK with live F454 + #0 (issue_629) and populated *#4*Z*0## (f461_live_2026-10-23). Cold-start #0 subordinate-zone setpoint hydration is tracked separately in #813 so this PR stays focused.
3. Make _polls_temperature a pure property (climate.py:397, 412-433) — right now it mutates _interface_poll_skip_logged and logs DEBUG during __init__ before the entity is attached to hass or the registry.
4. Resolve the self-contradiction in docs/getting-started/hardware-compatibility.md (L55-66) / scripts/update_gateway_profiles.py (L183-236), where Column 2 (Initial Discovery Subsystems Queried) still lists 4 and 16 on models whose Column 3 note says the sweep omits them, and add the missing MH201 / HL4684 notes.
5. Document in docs/configuration/services.md (L139-145) that myhome.sweep_bus (force: false) now sends zero WHO 4 frames on the gateways refusing *#4*0## (and does not trigger per-zone polls).
6. Back tests/test_startup_requests_per_firmware.py with real fixture frames (it currently loads zero trace files and uses synthetic Zone 3 frames despite the PR body stating trace frames are inlined there), and tighten the PR body's run6_event_session_ha_fixed_integration table entry (run6 is an evt-only rx capture and cannot by itself prove tx omissions).

(Separate trace-corpus schema, timestamp causality, issue_674 forensics, and fixture privacy-gate follow-ups are tracked in #814.)

[gdluck]

Thanks for the second pass; all six are on the branch (head d3be161, rebased on 089fef5, one commit per item).

1. MH201 (e4d8adc): added to GENERAL_ZONE_REQUEST_REFUSED on the issue_466 evidence: sweep frame 11 is *#4*0## and no zone frame follows it (only *#4*0#0*17## three times and zone 105's dimension 15), while the sibling trace shows zones 2, 3 and 4 (*#4*4*0*0241##, *#4*2*14*0100*3##, *#4*3*12*0090*3##; the three *#4*Z*0*TTTT## frames quoted in the review are not in that file, the zones are proven by the frames above). The 3.6.27 fresh-discovery log carries no WHO 4 traffic at all (the plant fixture has no climate), so it is cited as neutral. runtime_behaviour.md L15 / L37, climate.md, the b16 note and ("MH201", False, True) are updated; the tests that used an MH201 as the default sweep model drop *#4*0##, and the delegated-WHO 4 follower case moves to an MH200N so it still proves the forwarded path.
2. Comments (3c2f5c9): *#4*Z## is expanded into the four dimension queries on the SCS gateways (F454 + #0 in issue_629 answers *#4*1##..*#4*15## in about 1.3 s; the F455 multiple timeouts #674 10 s hang was an F455 on actuator-only channels); the *#4*1*0## NACK on 11 of 12 images is the bus TX timeout of an unpopulated zone (live F461 zone 1 NACK after 2.07 s, rows 374-376) while the populated zone 2 answers in 119 ms (rows 521-525); _poll_answered stays off "until a command-session capture of *#4*Z*0## on an SCS-only gateway exists"; *#4*Z*0## kept; feat(climate): hydrate cold subordinate zones via one-shot *#4*Z## once live #0 central unit is confirmed #813 named for the #0 cold start. The body says the same.
3. _polls_temperature (a4abfe6) is a pure property; the one-shot DEBUG line and the flag live in async_update, placed before the fan / humidity early return so a humidity zone behind an F422 on an F454 still logs once while sending only *60## (an independent second-opinion review caught that ordering). Tests: nothing logged by __init__ or by reading the property, one record on the first update, none on the second, F455 interface zone silent, the humidity case.
4. Compatibility table (2af4e8b): split rather than trimmed, because calibrate_discovery_with_ownd recomputes column 2 from the OWNd profile and verify_profiles_against_oracle requires equality with it. Column 2 stays "Profile Subsystems (advertised)"; a new column "Startup Sweep Requests (sent)" comes from a validated per-row sweep_omits (only *#4*0## / *#16*0*5## of an advertised WHO are accepted), and test_hardware_compatibility_table_matches_the_sweep compares it with _discovery_frames() for every row. MH200 and MH200N are separate rows, the MH201 note (refusal plus *#4*<WHERE>*0##) and an H4684 / HL4684 row are added; regenerated with sync_documentation.py --update.
5. services.md (da8c29b): default myhome.sweep_bus sends zero WHO 4 frames on F450, F454, F459, F460, F461, H4684, MH200, MH201, MH202 and MyHomeServer1 and triggers no per-zone poll; force: true sends them and the firmware drops them.
6. Tests (d3be161): tests/test_startup_requests_per_firmware.py loads f461_live_2026-10-23, myhomeserver1_live_2026-10-24 and the issue_466 MH201 sweep plus its sibling with the loaders the other trace modules use, and replays the captured frames (F461 zone 2 ack / zone 1 NACK / *#4*0## refused locally; an MH202 zone sends exactly the captured *#4*2*0##; MyHomeServer1 zones 35..70 on a MyHomeServer1 handler; the MH201 sweep frame 11 with no zone reply and its zone 4 polling *#4*4*0##); the synthetic zone 3 cases are kept and labelled. In the body, run6 is credited only with the bus-visible replies and the tx omission is attributed to the HA debug log of the live retest, quoting what HA logged. 70 cases (was 60); 15 of the new / changed ones fail on the previous head.

Full suite 4631 passed / 1 skipped, custom_components at 100 % statement coverage; ruff, typing ratchet, HA standards, docs sync and translations clean.

[GreenGrassBlueOcean]

@gdluck please rebase

[gdluck]

Rebased onto v2-phase1-architecture 5641f91 (after #792 and #815): the twelve commits replay with two resolutions in gateway.py (the #792 request tables and this PR's refusal tables now coexist, with _ZONE_REQUEST_MODELS keyed on all of them, and the SSDP HL4684 mapping folded into the base's _model_key) and one in climate.py (actuators and current_temperature_restored attributes side by side). Head b1c4716, mergeable; full suite 4752 passed / 1 skipped at 100 % coverage, ruff, typing ratchet, HA standards, docs sync and translations clean.

[GreenGrassBlueOcean]

@gdluck please rebase

[gdluck]

Rebased again onto v2-phase1-architecture 217b20e (after #793): the twelve commits replay on #793's merged resolver (its _model_key / KNOWN_GATEWAY_MODELS now also key this PR's three tables, _ZONE_REQUEST_MODELS unions all of them, and the SSDP HL4684 mapping is the one #793 brought), with one semantic reconciliation in the tests (13240f2): a humidity-refusing SCS-only zone (F450, F454, MH202) now sends its per-zone *#4*Z*0## temperature poll from this PR and still no *60##, so #793's "nothing is sent" cases move to the models that forward the sweep (F453, F453AV, MH200N) and the F422 humidity-zone case uses an F459. Head 13240f2, mergeable; 4855 passed / 1 skipped at 100 % coverage, ruff, typing ratchet, HA standards, docs sync and translations clean.

[GreenGrassBlueOcean]

All items from reviews 5642254816 and 5642844379 plus the #792 / #793 rebase reconciliations (13240f23) are verified against the firmware oracle and physical traces (f461_live_2026-10-23, myhomeserver1_live_2026-10-24, issue_466), and all 17 CI checks pass. LGTM!

[GreenGrassBlueOcean]

Following up on the MH202 010024 crash alert in own-firmware-oracle#63, we decoded the raw scs_bus_hex telegrams and disassembled libthermo.so / libplug_thermo_tr.so across all twelve emulated gateways (own-firmware-oracle#65, documented in OpenWebNet-Encyclopedia#52 and openwebnet-mcp#22):

1. Zero of the twelve emulated gateways actually route WHO 4 requests across an F422 interface ($06EC... / $06E4...):

   Frame	Gateways with bus_outcome=out	Decoded scs_bus_hex	Actual SCS bus effect
   *#4*2#4#01##, *#4*2#4#12##	H4684 020054, MH200 020102	$0499023000\r + $0499023012\r + $049902300E\r + $049902300D\r	Truncated at # (atoi) -> polls local main-bus zone 2
   *#4*2#4#01##	MH202 010024	(none — nack, crash)	libthermo.so parses trailing #4#01 as zone 4, actuator 1; thermo_open2scs+0x394 dereferences NULL -> coso SIGSEGV (*#4*2#1## crashes identically; 2#4#12## is rejected cleanly because 12 > 9)
   *#4*2#4#01*0##	H4684 020054, MH200 020102	$0499023000\r	Truncated at # -> polls local main-bus zone 2
   *#4*2#4#01*19##	H4684 020054, MH200 020102	$0499023013\r	Truncated at # -> polls local main-bus zone 2
   *#4*2#4#01*20##, *#4*2#4#12*20##, *#4*2#1#4#01*20##	F450, F454, F459, F460, F461, MyHomeServer1	$04A9023010\r / $004A9023010\r	Truncated at # -> polls local main-bus zone 2
   *#4*2#4#01*20##, *#4*2#1#4#01*20##	MH202 010024	$0499843110\r (2#4#12*20## -> silent)	Misparsed as local zone 4, actuator 1
   *#4*2#4#01*60##	F459, F460, F461, MyHomeServer1	$06D102030308000000\r / $006D102...	Truncated at # -> polls local main-bus zone 2
   *#4*2#4#01*11##	(none)	(none — 12/12 nack, silent)	Dropped on all 12 gateways

2. We opened #816 against v2-phase1-architecture to withhold unrouted WHO 4 #4#II polls across gateway.py (_send_zone_requests / interface_zone_request_refused) and climate.py (status, *11##, *60##) and guard dimensionless *#4*Z#N##. Inline notes below highlight the hunks in #794 affected by this so the rebase stays straightforward.

[GreenGrassBlueOcean]

H4684 (020054) and MH200 (020102) do not actually route *#4*2#4#01*0## or *#4*2#4#01## across an F422 interface: their WHO=4 parser reads atoi("2#4#01") == 2 and puts $0499023000\r on the main bus (polling local zone 2, identical bytes on *#4*2#4#12## as well; see own-firmware-oracle#65). Leaving H4684, MH200, MH201, F455, and Generic outside this set means:

• on H4684 / MH200, a zone 2#4#01 silently cross-polls local zone 2;
• on a Generic profile connected to an MH202 010024, supports_zone_status is True, so a zone 2#4#01 (01..09) sends *#4*2#4#01## and crashes coso with SIGSEGV.

Defaulting interface_zone_request_refused to True unless getattr(profile, "supports_interface_zone_requests", None) is True (as implemented in #816) avoids both cross-polling and the Generic -> MH202 crash while keeping the explicit profile opt-in.

[GreenGrassBlueOcean]

To match actuator_request_refused and humidity_request_refused (after #792 / #793), interface_zone_request_refused (and zone_temperature_poll_needed) should evaluate on self._who4_target() so WHO 4 secondary delegation and warm-standby failover consult the gateway that actually transmits the WHO=4 status request.

[GreenGrassBlueOcean]

Gating _interface_poll_refused on self._temperature_poll_wanted only withholds *#4*Z#4#II*0##, leaving three other #4#II paths open in async_update:

1. On F460, F461, and MyHomeServer1 (supports_zone_status=True), *#4*Z#4#II## still goes out (nack, silent on all three), tripping _poll_health into a false unresponsive repair issue after two polls.
2. On a Generic profile talking to an MH202 010024, *#4*Z#4#01..09## goes out and crashes coso (SIGSEGV).
3. Lines 671-676 below still send *#4*Z#4#II*11## (nack, silent on 12/12 oracle images) and *#4*Z#4#II*60## (truncated to $06D102030308000000\r / $006D102..., cross-polling local main-bus zone Z on F459, F460, F461, and MyHomeServer1).

Evaluating _interface_poll_refused independently of _temperature_poll_wanted and returning early in async_update when not self._central and self._interface_poll_refused (landed in #816) covers *#4*Z#4#II##, *0##, *11##, and *60## in one place.

[GreenGrassBlueOcean]

Worth updating this sentence (and the matching parenthetical in .github/release-notes/2.0.0b16.md line 24): in own-firmware-oracle#65, H4684 (020054) and MH200 (020102) do not forward #4#II across an F422 ($06EC / $06E4) — they truncate Z#4#II at # and emit $0499ZZ3000\r on the local main bus, so routed Z#4#II zones should wait for spontaneous reports across all models unless a profile explicitly sets supports_interface_zone_requests=True.