docs(debug-journal): fold in the 2026-10-07 queue slice (3 candidates); mark GH#5349 fixed by PR #5360; note #5392/#5403 refinements; fix drifted should_replace_plan line cites

[springfall2008]

Automated journal flush (the /journal-update daily flow) — folding the 2026-10-07 triage queue slice into tools/debug-journal.md, with every candidate re-verified against main (82bee43, v9.3.6) before anything was written. Nothing was left queued: the queue held exactly 3 candidates (limit 15), all 3 folded in; none dropped.

Folded in

1. GH#4445 — plan-table HoldChrg classified from target, not the real charge limit — added a symptom-table row: the State cell and status text classify from charge_window_best[n]["target"]; discard_unused_charge_slots() stamps target = limit only when predicted SoC sits at/above the limit at both window start and end, and otherwise stamps the predicted end-of-window SoC — so a rate-limited charge that cannot complete renders as HoldChrg -> <predicted SoC> and get_charge_export_text() says "charging to N%". Freeze classification already reads the limit (is_freeze_charge(limit)), so the two can also disagree; a rising charge makes completing charges label correctly by accident. Verified on main by reading the symbol chain plus a scratch unit replay on the publish_html_plan() path; no PR references Plan shows 'Hold Charge' for windows that are actually issuing a full-rate charge #4445 and no commit touches the classification since the issue was filed. Kept the mechanism, the fixture gap (test_plan_why_reason.py builds windows without a target key, so only the fallback path is exercised), and the fix direction as one line; dropped the scratch-test narrative and reporter details (per-issue narrative stays in the issue).
2. GH#5413 — planned dispatches on the Octopus entity are empty by design for an Ohme charger — added to the Octopus row: BCD's get_intelligent_features() special-cases OHME to planned_dispatches_supported = False (OHME is not in FULLY_SUPPORTED_INTELLIGENT_PROVIDERS), and the dispatches coordinator then clears planned dispatches outright (log: Clearing planned dispatches due to not being supported for provider), keeping completed and started. Octopus's API has carried Ohme Home Pro sessions as smart-charge planned dispatches only since ~29 Sep 2026 (BCD FR Issue where clipping removes a charge hold #1893, still open). Consequence: octopus_intelligent_slot → the BCD *_intelligent_dispatching entity on an Ohme-scheduled setup buys completed-slot tracking but zero forward car-slot visibility — and an empty planned_dispatches attribute is the gate's doing, not evidence Octopus holds no dispatches for Ohme. Predbat-side reads re-verified on main (fetch.py): completed merge into octopus_slots unconditionally, planned gated on plugged-in/ignore-unplugged, started diagnostics-only per IOG future planned dispatches treated as cheap home rate before EV charging earns the slot #4516 Stage 1 — so the IOG extra car charging slots still not holding the off-peak rate #5413 thread's fix shape (merge completed from Octopus with planned/started from Ohme) is the real fix. BCD source re-verified live today.
3. GH#5434 — car-conditional write/replan churn — folded in two places. New symptom row ("Too many inverter register writes / replans while the car charges"): the hour-bucket write-counting technique (grep "write_and_poll_value: Wrote" → hour buckets → sort | uniq -c, version A/B dated from the currently running banners + restart markers), and the five churn mechanisms — dynamic-load flap → forced replan; watch_list event-entity refires (an event entity has no comparable state, so every refresh bump is a full replan); flat-band limit churn under metric_min_improvement_plan (the GH#5330 tie surface); low-power mode doubling each limit flip into charge_rate + timed-current writes; window wipes. Ruled out Don't trust Intelligent slots the car won't need (consider_full) #5403 with the grep-for-its-log-line instruction. And the Best charge window = prefilled-incumbent trap appended to the attachment-forensics bullet: the recompute block resets the windows and prefill_charge_limit_best() runs before the line prints, so it shows the incumbent limits on the reset windows; the optimised plan appears only in the later Raw/Unclipped/Filtered lines, and the adoption decision's metric comparison re-scores both plans on the same cycle's inputs. Dropped: the reporter's counts/brands and the not-verified fix ideas (left with the maintainer on the issue).

Corrections to existing entries (merges since the last flush, de3fea8 2026-10-03)

• Sunsynk/Deye row — PR fix(sunsynk/deye): dont expire control_active after 15 minutes (fix #… #5360 (c246275, merged 2026-10-05, v9.3.6) split the restore caches — the row's PR fix(sunsynk,deye): persist control_active across a restart #5150 wording is superseded. applied_payload keeps the 15-minute change-detection bound (SUNSYNK/DEYE_RESTORE_MAX_CONTROL) while control_active moved to its own ownership cache under an 8-hour bound (*_RESTORE_MAX_CONTROL_ACTIVE), and run() calls _reconcile_control() on every tick that owns the inverter — so an HA restart inside 8h restores ownership and a dropped payload is followed by a real rewrite. GH#5349's "holds lost for hours after an HA blip" is fixed for outages inside the 8h bound; the row keeps the pre-fix(sunsynk/deye): dont expire control_active after 15 minutes (fix #… #5360 mechanism (≤ v9.3.5) for older logs, where the tail-end info line control ownership cache is ... minutes old (limit 480), requiring a fresh write is the expiry tell. Verified by reading the merged code and the new constants.
• PR Plan fails to schedule house battery to charge at same time as car since 9.3.4 (overnight or IOG extra slots) #5392 (445f547, v9.3.6) tightened io_adjusted flagging — refinement, recorded in both entries that describe it (the fix(octopus): model a vanishing IOG dispatch for octopus_intelligent_slot users #5304 symptom row and the Octopus row's GH#5012 passage): rate_add_io_slots() flags/re-prices only minutes a dispatch lowers by more than a penny (IO_RATE_TOLERANCE), so sub-penny dp2-rounding differences no longer mark or re-price the fixed off-peak minutes.
• Plan-stuck row line drift (found while verifying candidates, not merge-driven). The row cited should_replace_plan() at plan.py:1103 and a fragmentation tie-break at plan.py:1114 — on main those lines are unrelated code (a docstring and a comment), and the fragmentation branch has sat inside should_replace_plan() since PR Break near-tie plan comparison toward the less fragmented plan #4301 (2026-07-26): it favours the less fragmented plan on a near-tie and adopts a cost-neutral less-fragmented plan outright. Dropped the dead line numbers and rewrote the wording to the current semantics.
• Also added a one-liner to the car-charging row for Don't trust Intelligent slots the car won't need (consider_full) #5403 (v9.3.6): with octopus_intelligent_consider_full on, future dispatch minutes no car can use are cut too, with a replan trigger logging Octopus Intelligent: dispatch minutes no car needs changed to N — grep for that line before reaching for it on a churn report (from the IOG charging - Multiple inverter setpoint writes when charging car #5434 triage's ruled-out note).

Quality gate

Hooks pass (markdownlint-fix auto-formatted the file once and the second pass is clean); the chained quick suite was hard-killed at the 10-minute Bash cap mid-run on this machine (documented failure surface for docs-only flushes — a docs-only diff cannot affect tests, and tests running at all proves the hooks exited 0); targeted basic_rates + units passed.

Do not merge — a human merge is the review gate on this file.

[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.