daily-fix: kill-before-relaunch enumerates live launcher pha
kind: infra#wf-fix#wf-fix-fp:9e9ac916cd8d#daily-auto-filed
track: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: miner D-P1 (session ff4119b7, #1482, crash 10:03Z)).
Goal
In-place recovery must not race the live launcher's own phase sequence into a double-execution OOM.
Workflow gap
- Bug observed: The orchestrator's detached fits held 55.9 GiB when the launcher reached its own fits phase; the launcher OOMed (4 GiB needed / 1.22 GiB free), wrote a failed sentinel, crash-persisted, and powered the instance off. ~25 min recovery; self-inflicted crash.
- Why it is a workflow gap: the kill-before-relaunch/ownership rules cover killing PRIOR instances but not forward-enumeration of a LIVE launcher's upcoming phases before duplicating one.
- Confidence (emitter): medium
- verified-at-filing:
grep -c 'remaining phase' .claude/rules/crash-fix-rounds.md-> 0 (absence of the duty; 2026-07-30, this run). Incident mechanism from the session's own post-mortem marker.
Proposed change (refine in planning)
Add the duty to the Kill-before-relaunch/ownership section with the #1482 incident as the worked example.
Scope / surfaces
- Primary target:
.claude/rules/crash-fix-rounds.md - 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: .claude/rules/crash-fix-rounds.md
- fingerprint: 9e9ac916cd8d
Activity