docs(debug-journal): fold in the 2026-09-30 queue slice (2 candidates); mark GH#4940 fixed by PR #5308, GH#5294 by PR #5295, add the PR #5297 write-retry policy

[springfall2008]

Automated PR prepared by the triage bot's /journal-update flow; a human merge is the review gate on this file.

Queue slice: 2 candidates taken (limit 15), 0 left queued — the queue held only today's pair. Every claim was re-verified against origin/main at e652b2b1 (v9.3.3) before folding, and the journal was re-checked against what merged since the 2026-09-29 fold (PR #5296): PRs #5295, #5297 and #5308 overtake three existing entries, all corrected below.

Candidates processed

Folded in — GH#5300, Tesla reserve reset races a service hook on the reserve register (journal-queue/5300-tesla-reserve-race.md) → appended to the HA write/verify row. Verified on main: the reset block (if self.set_reserve_enable and resetReserve: inverter.adjust_reserve(0)) has no isCharging guard — one sharpening of the candidate: resetReserve starts True only when set_charge_window or set_export_window, not unconditionally; the clear branches are the freeze / hold-charging / car / iBoost holds; adjust_reserve() floors at reserve_percent and writes through write_and_poll_value(); the shipped-template immunity was confirmed in templates/tesla_powerwall.yaml (has_reserve_soc: False, the Note: Inverter does not support reserve disable in fetch_inverter_data()) and the dummy-arg path (create_missing_arg/create_entity); the "Current Reserve is X%" / "Wrote X to reserve" discriminator kept as the real-vs-dummy wiring check, plus the dump-config tells (expert-mode-hidden switch reads value: null yet resolves True; unknown args: keys ⇒ hand-built config). The candidate's line cites had drifted on main (inverter.py:3237 / :2006-2011 / :717-719) and were replaced with symbols.

Folded in — GH#5303, reset_inverter() writes a 100% target on target-SoC-only inverters (journal-queue/5303-reset-inverter-100-target.md) → new row in the symptom table. Verified on main: the 100% "safe mode" write with no inv_has_charge_enable_time consult in reset_inverter()/adjust_battery_target(); the 13 "reset_inverter": True CONFIG items (re-counted: 13, plus 2 reset_inverter_force items for mode select / set_read_only), None→value at startup counting as a change (userinterface.py); mode gating in fetch.py, where the startup reset-100 fires in the charge-controlling modes and the SoC-only modes only via the forced paths; the same-cycle corrections (plan target near a window, and the steady-state 0 write for has_charge_enable_time: False inverters outside a window); standing cases (entering Read-Only via the force gate, or Predbat stopping right after); the other 100% writes (calibration branch; self_test() under the INVERTER_TEST dev constant); the identically named reset_inverter(my_predbat) fixture helper (tests/test_infra.py:629); and the templates/victron.yaml 404 (re-confirmed 404 against the raw URL; no such file in templates/). Drifted line cites in the candidate were replaced with symbols (e.g. const.py:103→110, inverter.py:1884→1897, inverter.py:1884-1885→1897-1898).

Existing entries corrected (with the merge that invalidated each)

• Solis row — GH#4940 marked fixed by PR fix(solis): plan charge and discharge at the inverter's own separate limits #5308 (merged 2026-09-29, v9.3.3). The entry described the asymmetric binding (battery_rate_max → max_charge_power, max_discharge_power bound to nothing) as "still live". The merge publishes a battery_rate_max sensor holding the larger of the two per-direction limits and binds inverter_limit_charge/inverter_limit_discharge with overwrite=False; mechanism kept for pre-fix(solis): plan charge and discharge at the inverter's own separate limits #5308 logs. The companion GH#5187 sentence in the same row and the "charge or discharge rate pinned far below the hardware's" symptom row were updated with it.
• GE Cloud row — GH#5294 marked fixed by PR fix(gecloud): re-read the device list hourly so added or removed hardware is picked up (#5294) #5295 (merged 2026-09-29). The entry's "a restart is currently the only mechanism" for added/removed-device discovery no longer holds: refresh_devices() re-reads the device list hourly (DEVICE_REFRESH_SECONDS) and select_poll_devices()/apply_devices() rebuild what is polled and drop removed devices' stores. The refresh_discovery()-is-not-device-re-discovery naming trap and the 5-day last_updated skip survive the fix and stay as caveats.
• HA write/verify row — PR fix(gateway): stop re-sending hub writes that never verify #5297 (merged 2026-09-29) made the write-retry ladder per type. The verify-window arithmetic entry now also records INVERTER_DEF[type]["write_max_retry"]/["write_backoff"] (only GWMQTT deviates: 3 attempts, 10s apart, and a 2-failures → one write per 5 min degraded state until it verifies; all other types unchanged), and the "hundreds of identical didn't complete" symptom row gained the GWMQTT-only exception to its "indefinitely" reading.

Dropped

None — both candidates survived verification, and neither duplicated coverage (the #5300 entry is placed beside its #5094/#4165/#4641 sibling but the reserve race itself was not in the journal; #5303 is a new symptom mapping).

Quality gate

All 14 pre-commit hooks pass on the change (cspell and markdownlint-fix included; no dictionary additions needed). The run_all --quick half of ./run_pre_commit cannot pass locally during the 00:00–01:00 BST hour: test_teslemetry_local_weekday_follows_the_base_clock asserts the host's datetime.now().weekday() against a UTC-based fallback — the crash documented in this journal's 2026-09-22 note. Reproduced bare at 00:12 BST (AssertionError at tests/test_teslemetry.py:2789) and confirmed unrelated to this docs-only change: under TZ=UTC ./run_all --quick the whole quick suite is green — "All tests passed (4 slow tests skipped, total time: 73.65s)".

🤖 Generated with Claude Code