docs(debug-journal): fold in the 2026-10-05 queue slice (11 candidates); mark GH#5399/#5401, GH#5402/#5405, GH#5349/#5360 and GH#4832/#4956 fixed; add the debug-enable-engine-switch symptom row and the TEST_REGISTRY-name trap

[springfall2008]

Automated journal flush — this body and the fold-in below were produced by the triage bot's daily /journal-update run (2026-10-06 flush, queue 2026-10-05 slice). A human merge is the review gate on this file.

Queue slice: first 15 candidates in filename order from journal-queue/ (12 files present, one empty — 11 candidates read). 6 left queued: none — the whole queue was within the slice; nothing left over.

Per-candidate outcomes

Candidate	Outcome	Why
1172-car-ready-time-24h-cap.md	Folded into the Car-charging row	Verified on main: plan_car_charging() clips every cheap window at the next ready time (end = min(window["end"], ready_minutes), past-time wrapped +24h) so the non-IOG car plan horizon is [now, next ready time), while low_rates already covers the full rate horizon; no select option disables the deadline; IOG cars bypass it. GH#1172's remove-the-deadline ask and GH#3890's date-dropdown ask are complementary/opposed completions — link both before closing either. All three enabling facts re-verified at 82bee43e (plan.py/fetch.py/const.py symbols unchanged); no merge since the triage touched these paths.
3017-plan-details-car-charging-sentence.md	Folded into the Car-charging row; corrected an existing entry	short_textual_plan()'s "Your car is currently charging." fires off car_charge_slot_kwh(minutes_now, minutes_now+5) > 0 and never consults car_charging_now (rest of the section likewise plan-derived) — folded next to the row's existing car_charge_slot_kwh sentence. Separately, the candidate's superseded flag on row "Predbat charged the car at peak..." was verified and applied: PR #5251's commit is contained in git tag --contains from v9.3.0 on, so that row's "(merged 2026-09-26, unreleased at the time of writing)" is corrected to "(merged 2026-09-26, released in v9.3.0)".
3034-octopus-charge-limit-additive-vs-absolute.md suspected (corroborated in-thread only, not from the integration source); the %-to-add mechanism itself is code+log-verified and folded. Separately re-verified that PR #5403 (consider_full car-need gate, merged the same evening) changes nothing with consider_full off: config.py's only change in the window is Wallbox args, and the new octopus_surplus_minutes()` work only runs with the switch on — folded as a #5403 rider so the next reader knows consider_full-on now also withholds the cheap rate from minutes no car needs.
5399-ohme-car-plan-has-no-executor.md	Folded into the Ohme row, as fixed	The finding was the direct basis of PR #5401 ("take the car plan from Ohme's own schedule when Predbat is not controlling the charger", merged 2026-10-05, unreleased — last release v9.3.5): the session slots become the car plan via the dispatch path, labelled charger schedule, no cheap rate, re-priced by reprice_charger_schedule_slots() on this cycle's rates; mode re-decided every poll. Entry folded as fixed-on-main with the pre-#5401 mechanism kept for older logs.
5400-set-reserve-min-display-lags-register-floor.md	Folded into the reserve-ratchet symptom row; corrected an existing entry	The display-only gap (register floor reaches reserve_percent only, set_reserve_min raised one-way from battery_min_soc; GE Cloud publishes entity min without binding battery_min_soc) folded into row "Reserve keeps ratcheting..." next to the knob family. The candidate's stale-journal find was re-verified independently and applied: #4832 is closed and PR #4956 is merged (2026-09-06, fixes #4953) — row "Plan shows a flat hold at the reserve..." no longer says the GE Cloud flavour is open.
5402-ohme-slot-owner-tariff-only-detect.md	Folded into the Ohme row, as fixed	PR #5405 ("leave the Intelligent car slots to Octopus when the Intelligent device is the car", merged 2026-10-05, unreleased) implements exactly the proposed device-check fix: octopus_intelligent_wanted() now consults the Octopus component's live non-suspended devices (is_ohme_charger() = is_charger + provider "ohme"); Octopus driving another device leaves the slots to Octopus and ohme_control stands down; explicit flag still wins. Folded as fixed-on-main, pre-#5405 signature kept.
5404-octopus-weekend-happy-hour-booking.md	Folded into the Octopus row, with one precision fix	One claim corrected against the tree: available_events is read — by the Power Down auto-join loop (join_octoplus_power_down_session_event + deprecated fallback) — so "read nowhere" was wrong; what does not exist is any Power Up/Happy Hour booking half (WHH entries are skipped at the source — cannot be joined through the API — and Power Up entries ride in available_events at 0 octopoints past the threshold guard). Folded as enhancement-not-bug with the implementation template and the 48h-horizon trap.
5408-debug-enable-engine-switch.md	New symptom row + kernel-trap extension	Verified on main: kernel_supported() requires not save and not pred.debug_enable and kernel_handle != 0 — debug on routes every trial prediction to the Python engine (deliberate, since #4169); 0 scan_soc_range count over a plan = kernel served it, non-zero = Python engine; committed libraries load from the module dir with no installer. New symptom row written ("Turning debug_enable on changes the plan..."), and the stale-kernel trap now carries the 0-scan-count tell. Kept the two loose ends flagged as loose ends (the predbat.py docstring still calls the Python path "more accurate but far slower" while kernel_supported()'s docstring says the opposite; no user doc) — flagged as maintainer calls, not silently asserted.
5408's loose-end re-check	(part of the row above)	The candidate's predbat.py:869-875 pointer had drifted; re-anchored on the _debug_enable_auto_scope() docstring (prebat.py:866-871) where the contradictory wording still lives. Also folded the 2h auto-disable (DEBUG_ENABLE_MAX_HOURS, since v8.53.0) as the bound on the switch.
5412-octopus-intelligent-device-freeze-after-tariff-change.md	Folded into the Octopus row, still live	Re-verified every link on main at 82bee43e: tariff-gated early return in async_update_intelligent_devices() → pruning never runs off-IOG; automatic_config() wires tariff-blind; cache round-trips the devices; self.octopus_slots rebuilt from frozen entities every cycle; num_cars raise-only; the automatic_config() clobber of a manual octopus_intelligent_slot still overwrites on the first run after restart (post-#5405 only clearing respects foreign wiring). Ohme-side stand-down still blocked by charger_slots_wanted() (the #5401 schedule mode stays blocked while the freeze lasts); #5405's car_slots_released_to_us() treats the not-refreshed device list as a live constraint by design. #5405 does not close #5412.
5413-iog-dispatch-stamp-clears-on-session-end.md	Folded into the Octopus row	Re-verified on main: completed dispatches merged unconditionally (fetch.py), dispatch_billed_off_peak() exempts completed, history re-render derives past minutes from the current tables (history_to_future_rates(self.rate_import, ...), today_cost() re-prices today-so-far). Folded as the completed-dispatch counterpoint to the row's existing #4946/#4957 AWAY-relabel mechanism; open #2785 (PR #3340 parked) cited as the persistence gap.
5415-iog-hold-for-car-empty-slot.md	Dropped — empty file	The queue file is 0 bytes; nothing to verify or fold. The underlying triage (IOG hold-for-car, referenced in the bot's memory as gated-no-theory) needs re-queuing by the triage run if it had a finding.

Existing entries corrected (merge-staleness pass)

• "Predbat charged the car at peak rate..." row — "car_charging_plan_smart default flip released in v9.3.0" (was "unreleased at the time of writing"); PR fix(car): default car_charging_plan_smart to on (#5237) #5251 verified contained in v9.3.0 via git tag --contains.
• "Plan shows a flat hold at the reserve" row — "GE Cloud does not bind battery_min_soc, so the planner models a floor the device rejects #4832 is the GE Cloud flavour, open" → closed; PR fix(inverter): model the reserve floor the register will accept, not just set_reserve_min #4956 merged 2026-09-06 (fixes Reserve: planner models set_reserve_min, not the device's published register floor (#4828 clamps only the write) #4953, the reserve-register modelling). Verified via issue/PR pages (gh api is denied on this session, so via WebFetch of the public pages).
• Sunsynk/DEYE row — PR fix(sunsynk/deye): dont expire control_active after 15 minutes (fix #… #5360 (merged 2026-10-05, unreleased, fixes GH#5349) split the restore caches and bounds: applied_payload keeps 15 min, control_active now 8 h (*_RESTORE_MAX_CONTROL_ACTIVE), legacy combined-cache migration included; the restore log lines split into "control payload cache ... forcing a rewrite" / "control ownership cache ... requiring a fresh write". The GH#5138 quoted line updated to pre-fix(sunsynk/deye): dont expire control_active after 15 minutes (fix #… #5360 wording, and the GH#5349 entry cut to fixed form with the mechanism kept.
• Wallbox row — noted the component landed on main in PR feat(wallbox): Wallbox EV charger cloud component #5410 (merged 2026-10-05, unreleased at the time of writing).
• Car-charging row — added the PR Don't trust Intelligent slots the car won't need (consider_full) #5403 consider_full rider (car-need gate: octopus_surplus_minutes() withholds the cheap rate from minutes no car needs, with consider_full on; behaviour unchanged off, so no existing claim invalidated).

Verification method

Every folded claim was re-verified against current main (82bee43e, v9.3.6-pending) by reading the named symbols on the working tree, and against the two OHME/Octopus PR merges (#5401/#5405) by reading their commit messages and diffs (git show); GitHub state checked via public pages where gh is denied. Suspected/corroborated-only parts stayed marked. GitHub state for #4832/#4956 checked via public pages (gh is denied on this runner; the candidate itself verified via gh).

Gates

• pre-commit hooks: all pass (markdownlint auto-fix applied then passed; one cspell catch fixed — a prebat.py typo, reworded rather than dictionary-added).
• Quick suite: the full run aborted mid-suite at test_teslemetry_local_weekday_follows_the_base_clock — the documented 00:00–01:00 local/UTC weekday-divergence trap recorded in this journal's own traps section (fifth sighting; same abort signature as 2026-09-20, 2026-09-22, 2026-09-25, 2026-09-28 and 2026-10-05); everything before it was green. The remaining ~112 quick-registered modules after teslemetry were then run explicitly (--test <name> per module, in registry order) and all passed (50.6s), so the suite is verified for this docs-only diff modulo that one environmental module — no code is touched by this change.