docs(debug-journal): fold in the 2026-10-11 queue remainder (5 candidates); mark GH#5471 fixed in PR #5475

[springfall2008]

🤖 Automated journal-update run for 2026-10-12 — this PR was opened by the triage bot; a human merge is the review gate on tools/debug-journal.md.

Queue: /Users/treforsouthwell/predbat-triage-bot/journal-queue with a 15-candidate limit; the directory held 5 candidates (all from the 2026-10-11 triage round), so this PR folds in the whole current queue — 0 left queued.

Per-candidate outcomes

Candidate	Outcome	Where / why
5483-predheat-load-forecast-wiring-exists.md	Folded in (condensed)	Predheat row: the wiring exists — predheat.heat_energy$external feeds the plan's forward load, LoadML is additive, not exclusive (fetch_extra_load_forecast() sums ML + every load_forecast entity; load_forecast_only only zeroes the historical component), double-count caveat via car_charging_energy. Verified end to end on d7d8c607 (predheat.py publish path, fetch_extra_load_forecast, get_from_incrementing, predbat.py instantiation, both doc pages). Fold-in note: one candidate claim drifted — the load_forecast_only variant is documented in docs/apps-yaml.md only, not both pages as the candidate said.
5485-compare-chart-bars-are-recorder-rows.md	Folded in	Compare row: a chart bar group is one HA-recorder row — the chart is rebuilt per render from recorder history of predbat.compare_tariff_<id> (get_history_wrapper + history_attribute(daily=True, first=True) keeps the first state per local day), the nightly run writes once/night (publish_only() suppresses), a render-time "today" point always renders, so missing groups map to recorder/nightly-run causes and the Avg7 (Nd) column is a no-files diagnostic. All symbols verified on d7d8c607 (web.py chart path, history_attribute day bucketing, predbat.py midnight gate, const.py, history_daily_at_hour).
5489-update-entity-skip-hardcoded-auto-update.md	Folded in	GH#5374 symptom row: skip/hide on update.predbat_version starts at the hardcoded auto_update: True (userinterface.py, deliberate since #727); HA's skip button blocks on exactly that attribute and not on the no-unique_id limitation (the two are separate); update.skip is captured but update_event()'s skip branch is a stub; any fix must persist skipped_version Predbat-side (Storage) and republish it. Verified on d7d8c607 (userinterface.py update branch/listeners/stub, github.py publish, config.py item).
5490-import-limit-ratings-only-no-planner-cap.md	Folded in	New symptom-table row: there is no import-side clamp in the planner — tree import_limit is per-inverter discovery ratings only (sunsynk/fox/deye), no config.py setting, prediction clamps export only (prediction.py); maps to "no such setting exists", with the recorded direction = draft PR #5486 + the GH#1199 design note (GH#5490 umbrella, GH#2449/#1199 family; car-planner ranking is the GH#5384 entry). Re-verified on d6985b6d (grep: ratings-only; no config item; prediction.py clamp). PR #5486 re-checked today: still draft, proposes diagnostics-only with no enforcement.
5492-...-partial-response.md	Folded in	Octopus row: KT-CT-4340 partial GraphQL 200s come back as data (ignore_errors=True + "return data whenever present" = the caller can't see errors), the or [] turns a blip's null into a real [] stored over the cache (slots/cheap rate vanish one cycle, 2 extra replans; the issue's log measured 7/230) — PR #5250's fallback covers only None; fix = route the null into it. Kraken contrast kept (kraken returns None on any errors, cache survives). Verified on d6985b6d (octopus.py:2401/2409/2496/2498/2554/2604-2620 area, kraken.py:375-378, Test 14).

What was dropped

Nothing dropped — all five candidates passed verification. One claim was corrected at fold-in (5483's doc claim, above). The 5492 candidate's aside on the charging-now rate ratchet (car_charging_rate raised to max(Octopus kW, configured)) was not folded because the Octopus row's GH#5438 entry already covers that mechanism — no second entry needed.

Journal re-check vs merges since the last journal PR

Only one commit merged after yesterday's journal PR (f20ddaf, 18:54) — PR #5475 (d6985b6, 19:40), which fixes GH#5471 exactly as yesterday's entry said "fix in flight". Corrected:

• Gateway row, GH#5471 — cut down to the fixed-in-PR-fix(gateway): keep configured inverters when the hub cannot read them (#5471) #5475 form; mechanism kept for pre-fix(gateway): keep configured inverters when the hub cannot read them (#5471) #5475 logs; "partial-outage scenario uncovered" replaced by "partial-outage tests now cover it" (the PR added ~1000 lines of gateway tests); the "fix trap" (re-run gating) is kept as the rule the merged fix itself honours.
• Everything else the journal claims about areas touched since 2026-10-08 checked out against d6985b6d (gateway.py rebased code, octopus.py byte-identical across the range, docs commits touching no journal-claimed behaviour).

Quality gate

Pre-commit run per coverage/run_pre_commit (hooks stage): markdownlint, cspell, ruff, black, export-limit checks, triage daemon tests — all green, no dictionary additions needed. The merged gateway fix's own tests ran in that PR's CI; no code changes in this PR.

🤖 Generated with Claude Code

[Copilot]

Copilot wasn't able to review any files in this pull request. Check if the Files changed in this pull request are included in default exclusions.