workflow-fix: launchers atomically refresh the pid file on every relaunch (stale-pid poll target)

Overview / Motivation

Auto-filed by the workflow-fix protocol from the #813 v5 relaunch incident (watch-session orchestrator; filed at the #813 park).

Goal

Every launcher/relaunch of a pod-side or VM-side detached workload MUST atomically rewrite the pid file (/workspace/logs/issue-<N>.pid or the VM analogue) with the NEW pid before the poller's next read; a relaunch that leaves a predecessor's pid in the file is a launch-contract violation.

Workflow gap

• Bug observed: #813's run-4 relaunch (2026-07-02 00:31) left run-2's stale pid (6267) in /workspace/logs/issue-813.pid, so the poll chain would have watched a dead process; a corrective run-5 launch (00:39) was needed solely to fix the pid file.
• Why it is a workflow gap: pod-side-reporting.md specifies the sentinel/[phase=] contract but not the pid-file refresh-on-relaunch invariant; the SKILL.md launch snippets write the pid at first launch only.
• Confidence (emitter): high

Proposed change (candidate diff sketch — refine in planning)

• .claude/rules/pod-side-reporting.md: add a "pid-file contract" bullet: launchers write the pid file ATOMICALLY (tmp+rename) on EVERY (re)launch; a relaunch script includes the pid-file rewrite in the same command chain as the setsid/nohup launch (the #813 run-5 correction as the worked example); pollers treat a pid file older than the newest run-launched marker as suspect.
• Cross-check .claude/skills/issue/SKILL.md launch snippets reference the rule (pointer only; keep the full text in the rule).

Scope / surfaces

• Primary target: .claude/rules/pod-side-reporting.md
• Grep first: grep -rn "pid_file\|\.pid" .claude/rules/ .claude/skills/issue/SKILL.md scripts/poll_pipeline.py | head

Constraints / invariants

• Workflow-surface only; workflow_lint passes.

Provenance

• workflow_fix_target: .claude/rules/pod-side-reporting.md
• fingerprint: (wrapper)