docs(debug-journal): fold in the 2026-10-11 queue slice (10 candidates); mark GH#5050 and the GH#5221 event-day instance fixed in PR #5163, GH#5461 fixed in PR #5462, and the Victron template gap fixed in PR #5332

[springfall2008]

Automated journal update — the /journal-update flow folding the 2026-10-11 queue slice. 10 candidates: 10 folded (one of them rewritten as fixed), 0 dropped, 0 left queued. Every claim was re-verified against origin/main d7d8c607 before folding. Heads-up: ./run_pre_commit, pre-commit run --all-files and the node linters are permission-denied for the triage runner, so the exact hooks could not be run — the applicable ones were replicated by hand (trailing whitespace, end-of-file newline, dictionary case-insensitive sort order, and a token scan of every newly introduced term against the workspace dictionary; one genuine new telemetry key, pac, was added). pre-commit.ci runs the real gate on this PR.

Merge sweep — existing entries corrected

• GH#5050 low-rate-threshold row (symptom table): PR fix(fetch): exclude saving-session/Axle boosted minutes from automatic rate thresholds (#5050) #5163 merged 2026-10-09, first released v9.3.7. The row said "PR fix(fetch): exclude saving-session/Axle boosted minutes from automatic rate thresholds (#5050) #5163 (open)" and described only the planned fix direction. Rewritten: marks the fix and describes its actual shape (threshold stats from rate_minmax_excluding_saving() — saving-tagged minutes mapped back to their pre-event rate, downward-only capping, user overrides applied to the pre-saving snapshot, synthetic minutes with no pre-event rate ignored, rate_max/min/average untouched, compare/annual clearing the saving-minute state, pre-event import windows via find_low_rate_windows()/low_rates_tariff for the plan while sensors and the non-smart car plan keep tariff-only windows, and the new test_set_rate_thresholds.py module), keeping the pre-fix(fetch): exclude saving-session/Axle boosted minutes from automatic rate thresholds (#5050) #5163 signature for older logs.
• GH#5221 export-threshold row: the same PR fixed the event-day "ratchet to 99" instance — the fix's own test pins a Feature: Native Solar Clipping Buffer (Resolves #1206) #4036/Export threshold can hide whole periods from the planner: slot visibility, not cost #5221-shaped flat-export-tariff control reproducing the unfixed behaviour. The lowest = 99 init in rate_scan_window() is unchanged, so the overwrite-path caveat for future rescans is kept.
• Victron docs trap ("Battery charges to 100%…" symptom row): PR docs(inverter-setup): rewrite Victron section and add templates/victron.yaml #5332 (d7d8c607) added templates/victron.yaml and rewrote the docs section (2026-10-10); the "template does not exist (404)" claim is now past tense.
• Octopus row: the load_saving_slot() line cite (octopus.py:2786) had drifted (now ~3180); replaced with the symbol name per the journal's own convention.

Per-candidate

1. 5454 (iBoost solar + iboost_smart combo) — folded into the iBoost row: the docs-vs-code contradiction (docs/customisation.md "only applies when iboost_solar and iboost battery are both Off" vs the fetch.py gate ((not iboost_solar) and (not iboost_charging)) or iboost_smart, both verified on main), the grid-first combination mechanism, the iboost_rate_threshold -10 floor, the PR iBoost: optional demand-forecast-driven smart planning (iboost_forecast) #4756/Consider solar export value when scheduling EV charging #5384 pointers and the flat-rate test trap.
2. 5456 (HAHistory prune drops the state-in-force record → Predheat 20.0 °C) — folded into the Predheat row: the hourly EntityHistory.prune_history() cutoff drops the state-in-force record at the cutoff, so an unavailable/unknown blip as the oldest surviving record floors age to 0 and get_historical(default=20.0) subserves for 24 h; the blip-non-empty-store mechanism (the 2-minute incremental refresh only appends newer epochs); the shared-cache distinction (Predbat's own load path floors age use at 1 day); and the fix trap (the one-line retention fix fails two test_hahistory.py assertions encoding the current prune semantics). Still live.
3. 5458 (Power Down reward assumed unconditional) — folded into the Octopus row, sharpening the existing GH#2090 sentence: Octopus pays the reward only on export above the user's baseline (typical export at that hour), so a plan that already exports in the event hour models a fictional session value; the three-way report split (inflated total = per-minute crediting Account for baseline calculations for Saving Sessions #2090, "I exported but got no reward" = baseline, low-value bookings = octopus_saving_session_min_octopoints_per_kwh Octopus Saving Session value overstated: incentive applied to whole export window, not marginal reduction vs baseline #4893); and the persistence gap (events re-read each cycle) plus the unprobed BottlecapDave Baselines attribute pointer.
4. 5459 (Sigenergy docs entity-name split) — folded into the Sigenergy row: the docs' predbat_max_*_limit_action automations cite inverter-level names with the plant -ing suffix — matching neither the plant -ing pair nor the inverter non--ing pair — and the template binds battery_rate_max to the plant -ing name; fix in flight on PR docs(sigenergy): fix rate automation sensor names, re-apply mode on restart, stop zeroing import limit in freeze #5464 (verified unmerged against current main).
5. 5460 (Sunsynk battery-first idle clipping, closed dup of PV clipping protection: proactively create battery headroom using PV forecast/PV90 #4865) — folded into the Sunsynk row: the planner-shaped-then-re-pointed + battery-first-idle-charge mechanism behind "planned export didn't happen / battery charged instead", the clipped-energy accounting sites (clipped_today, the inverter_can_charge_during_export + inverter_support_feedin_first branch — symbols verified on main), and the reusable log-archaeology traps (grid power = −pac, battery sign conventions, ~15-min poll vs 5-min publish stale cache, etoday counters ratchet, spot-sample balance reconciliation fails).
6. 5461 (IOG low-load test) — rewritten as fixed: PR fix(car): low-load test needs live load power to be low too before cancelling an Intelligent dispatch #5462 merged 2026-10-09 (v9.3.7): the live load_power total must also be under the cut-off (no load_power → the 5-minute figure alone decides; not-low or no-reading → no evidence either way), and the cut-off factor dropped 0.9 → 0.85 (DYNAMIC_LOAD_CAR_LOW_FACTOR, verified on main). Durable mechanism kept: two-consecutive-low-window grace, ramp-window eligibility, confirm/cancel band asymmetry (0.85 vs the full threshold), the car_charging_energy-sensor strip skip; pre-fix(car): low-load test needs live load power to be low too before cancelling an Intelligent dispatch #5462 signature kept for older logs. The candidate's "byte-identical v9.3.5→main" claim is superseded by the fix.
7. 5471 (GatewayMQTT auto-config drops an unreadable inverter until restart) — folded into the Gateway row: partial visibility proceeds with the readable subset while _configured_inverter_serials records all serials so _needs_reconfigure() never sees the dropped device as new; the fix trap (a serial-set-membership re-run thrashes every frame — gate on the device becoming usable); Gateway MQTT: defer initial auto-configuration while the hub reports more dongles than inverters #5227's deferral does not cover "listed but unusable"; the green gateway suite proves nothing for the partial-outage scenario; fix in flight on PR fix(gateway): keep configured inverters when the hub cannot read them (#5471) #5475 (discovery/auto-config side).
8. 5478 (GatewayMQTT missing-field fallbacks fabricate 6000 W / unlimited export) — folded into the Gateway row: the or 6000 rate fallbacks, the proto-0 export-limit 99999 sentinel read back as 100.0 kW, literal-0 reserve_soc publishing, and _check_rate_caps() persisting the fabricated cap on the hub; or 6000 dates from the component's first commits (not a regression); fix precedent merge_non_null(); a green gateway suite proves nothing here (no missing-rating case).
9. 5479 (docs yaml code fences unvalidated) — folded as a new "Traps when investigating" bullet: nothing validates the fenced yaml in docs/inverter-setup.md (check-yaml matches only .yaml/.yml; verified against .pre-commit-config.yaml), the two SolaX script blocks broken since Feb 2026 by a wide neighbouring re-indent (PR Docs: Improve LuxPower Freeze Charge Exit Automation: Fix Trigger/Condition Logic to Handle Predbat Restarts #3389), the two break shapes (hard parse error vs silently-parseable variables: null), and the git blame -L technique.
10. In addition, one pre-existing malformed symptom-table row (the "Too many API errors" row missing its closing |) was repaired, and the merge-sweep also surfaced the "manual override partially ignored" class gaining a fixed instance (GH#5423, fixed by PR fix(plan): shrink, not grow, the source window in the export swap's partial combine (#5423) #5447) now noted in the web row.