daily-fix: watcher leaves filed infra tasks undispatched for
Highlight text on any card (body, plan, or an event) to anchor a comment, or leave a whole-task comment here. Mention @claude to summon a reply.
No comments yet.
Overview / Motivation
Auto-filed by /daily 2026-07-29 (problem sweep; emitting source: miners A-P1, E-P1, G-P1, J-P1 (4 independent groups), verified live by the orchestrator).
Goal
A filed, ripe, auto-dispatchable infra task must actually get dispatched — the watcher's proposed_infra_sweep is the documented ~10-min backstop, and an urgent-main-red task must not sit behind a day-long queue.
Workflow gap
- Bug observed: Ten filed infra tasks sat at
proposedwith onlyepm:createdfor 8-22h (#1823 filed 07:57Z by the urgent-park router with session_spawned:false; the router's later dedup hit against it posted a routed-record — 'deduped against #1823 ... session_spawned: false', #1800 events 2026-07-30T00:23:18Z — without dispatching it). Meanwhile main stayed red 40+h and every session's Step 9c gate re-classified the red (fixed tonight by /daily route-1). Later-created tasks (#1829, #1838, #1844) got sessions via their filers' direct spawn. - Why it is a workflow gap: the sweep is the backstop for filers whose spawn no-ops (cap full / daemon unreachable); if it does not drain — or drains oldest-first with a permanently full cap and no urgency ordering (probed: ids.sort() oldest-first, miner A) — filed==dispatched breaks and urgent reds live for days.
- Confidence (emitter): medium
- verified-at-filing:
task.py viewon the 10 ids -> each status proposed, 1 event (2026-07-30, this run); E-P1 probed the dedup branch at scripts/autonomous_session_watch.py ~L10212 (dedup_hit is not Noneposts the record, never dispatches) and~/.eps-autonomous/infra-drain-queue.jsonlast updated 2026-07-17.
Proposed change (refine in planning)
Three legs, planner to scope: (a) urgent-router dedup branch best-effort dispatches a never-dispatched proposed kind:infra dedup target (same daemon/cap gates as _urgent_wf_park_file_and_dispatch); (b) urgent-main-red tag gets priority (or a fast lane past the cap) in proposed_infra_sweep; (c) root-cause why the sweep never dispatched any of the 10 (cap accounting during the 07-29 wave of ~38 sessions? predicate miss on router-filed tasks? the stale infra-drain-queue.json?).
Scope / surfaces
- Primary target:
scripts/autonomous_session_watch.py - Grep the workflow surface for the pattern before editing and update every hit.
Constraints / invariants
- Workflow-surface only;
scripts/workflow_lint.py --check-askspasses; ruff clean on touched files.
Provenance
- workflow_fix_target: scripts/autonomous_session_watch.py
- fingerprint: 77c1ec5a75f0