fix: preserve live pipeline run state during crew reporting

[bluewandd]

What Changed

• Attribute read-only crew state to same-branch pipeline runs whose nonempty commit heads are not yet locally available, while retaining strict matching for teardown aborts.
• Prefer active matching run rows over terminal rows when resolving coarse run listings.
• Document the separate reporting and teardown attribution policies and add regression coverage.

Risk Assessment

✅ Low: The reporting-only attribution relaxation is confined to fm-crew-state while teardown retains the strict verified-head predicate for destructive aborts.

Testing

Reviewed the target diff, ran the two focused behavior suites, and exercised the crew-state CLI end-to-end against representative no-mistakes responses; both unresolved live-run scenarios rendered state: working · source: run-step, with no working-tree artifacts left behind.

Evidence: Crew-state CLI end-to-end transcript

Scenario 1: pipeline-owned validation reports a head not yet in this worktree; its own axi status remains working.
state: working · source: run-step · validating (fixing)

Scenario 2: runs list puts a stale failed row before the current unresolved running row; active row must win.
state: working · source: run-step · validating (background run)

Pipeline

Updates from git push no-mistakes

⏭️ **intent** - skipped

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

🔧 **Review** - 1 issue found → auto-fixed ✅

• 🚨 bin/fm-nm-run-lib.sh:78 - This shared ownership predicate now accepts any unresolvable head, but fm-teardown.sh uses it before axi abort --run and explicitly requires a verified branch+head match. A stale/invalid or GC-pruned parked run on a reused branch therefore passes attribution and is aborted during teardown. Keep the resolvable-head requirement for destructive callers; add a reporting-only compatibility predicate (or explicit mode) for fm-crew-state's pipeline-owned containment path.

🔧 Fix: Separate reporting-only run attribution from teardown safety
✅ Re-checked - no issues remain.

✅ **Test** - passed

✅ No issues found.

• git diff --check e836a7f28bc5100f517c1d672a56483754899add 40622263504b110ac729a57ea1351fc3c6dc8117
• bin/fm-test-run.sh tests/fm-crew-state.test.sh
• bin/fm-test-run.sh tests/fm-teardown.test.sh
• Manual bin/fm-crew-state.sh CLI exercise with a pipeline-owned unresolved head and with a stale failed runs-list row preceding a live running row.

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

[bluewandd]

Following up on this fix. The PR has all 13 checks passing and has been open since August 11. It prevents fm-crew-state from reporting a live pipeline-owned run as failed while keeping teardown\x27s strict branch/head safety. The branch is behind current main and needs a rebase before merge. Please review when you get a chance.

[kunchenguid]

Speaking as Kun's firstmate: first look on current main 85d6c7226c93 (#2953). Never messaged the captain.

class=corrective. Reporting-only attribution: a nonempty run head missing from the local object store stays reportable so a live pipeline-owned no-mistakes run is not mapped failed; coarse listing prefers an active same-branch row over a terminal row. Teardown keeps the strict fm_nm_head_matches_worktree predicate. Not leftover of landed #2953 (actor-split; main still uses strict crew-state matching). Related to open afternoon #3001 (same false-failed live NM run, broader: classify/watch wedge too). Help this PR; do not open a competing one.

VISION.md (inspected bin/fm-nm-run-lib.sh fm_nm_head_matches_worktree_reporting, bin/fm-crew-state.sh coarse active-over-terminal, bin/fm-teardown.sh comments-only strict abort, tests):

• One captain, one interface: aligns (false failed hid in-flight validation).
• Authority is explicit: aligns (no new grant; reporting variant must not authorize teardown).
• Scripts own the mechanics: aligns (resolvable vs missing head is exact).
• A restart is a non-event: aligns (live run remains inspectable).
• Delegation with a spine: aligns.
• The fleet outlives any vendor: aligns (binds to no-mistakes run listing / axi status).
• Scope: aligns.

This HEAD: 19d67428441f371ebf9677ed43a1cabc39e8982b. CONFLICTING / DIRTY, ahead 3 / behind 67.
Attestation c376f6e8d2c3f7dfe83d746ede7c4c1592aa7358 ≠ THIS HEAD.
CI green on this SHA but stale (runs 31737096580 / 31737096644 from 2026-08-13). Author follow-up 05:44Z asked for review and noted rebase need. Author body does not close an issue.

Not first-time: Behavior/Lint/CI actually ran on this PR. No workflow approval this pass. Greptile is not a merge gate.

Security: none. Read-only reporting relaxation; teardown stays fail-closed.

Overlap: bin/fm-crew-state.sh / bin/fm-nm-run-lib.sh with open #3001; bin/fm-teardown.sh with #2804 / herdr pair #2637/#2692 (this hunk is comments + a test that teardown does not abort an unresolvable head). Not a lock PR. Not spawn-freshen.

Land-eligible rec: NO. Captain-flag NOW: no.

This is waiting-on-author (rebase onto current main + HEAD-matching attestation; coordinate with #3001 rather than land both). Not a captain-decision. Not a merge I will recommend.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge because reporting is relaxed without weakening the strict ownership check guarding teardown aborts.

The changed reporting behavior is deliberately read-only, destructive teardown continues to require a resolvable code-identity match, and no concrete unacknowledged failure remains.

_{Reviews (1): Last reviewed commit: "no-mistakes(document): Document reportin..." | Re-trigger Greptile}