fix(bin): accept a task's next PR after its bound PR merges in fm-pr-merge

[karotkriss]

Intent

Fixes #5697

When one task produces several PRs, bin/fm-pr-merge.sh merges only the first; every later merge is refused with "error: task is bound to , not ", even after already merged through fm-pr-merge.sh itself, so an operator has to re-run bin/fm-pr-check.sh by hand before each later merge.
Once the task's bound PR has merged, fm-pr-merge should accept the task's next PR; while the bound PR is still unmerged it should keep refusing.

What Changed

• require_recorded_pr_identity in bin/fm-pr-merge.sh now accepts a task's next PR when the recorded pr= URL differs, provided that bound PR has already merged (proven by its recorded merge notification via fm_pr_poll_merge_already_notified); while the bound PR is still unmerged, a different URL is still refused. The bound URL is parsed in a subshell so FM_PR_* stays the new URL's identity for later callers.
• Updated the header comment to document the "already-merged bound PR" exception to the pr= binding rule.
• Reworked the test_distinct_merged_prs_keep_distinct_wakes test to rely on the confirmed-merge marker instead of hand-editing pr=/pr_head= out of the task meta before the second merge.

Risk Assessment

✅ Low: A small, well-bounded gate relaxation that follows an existing tested idiom, correctly implements both sides of the stated intent, and is covered by a behavioral test.

Testing

Drove the two end-user scenarios directly against the real bin/fm-pr-merge.sh through the test harness's run_pr_merge entrypoint: a task's second PR is accepted once the first PR's merge is notified, and a different PR is still refused while the bound one is unmerged. Both passed. Proved the regression by reverting only the fix, which made the second merge fail (refused as bound to the first URL), then restored the file and left the worktree clean.

• Live validation: ✅ go - 3 of 3 scenarios driven live against the product

Scenario	Result	Live	Evidence
After the bound PR merges, the task's next PR is accepted	✅ pass	live	test_distinct_merged_prs_keep_distinct_wakes drives real bin/fm-pr-merge.sh: merge PR/68 writes the notified marker, then merge PR/69 (pr= still bound to /68) succeeds and both leave distinct wakes
While the bound PR is unmerged, a different PR is refused	✅ pass	live	test_away_record_does_not_bypass_red_or_identity: pr= bound to /99 (no notified marker), merging /85 exits 1 with 'is bound to https://github.com/example/repo/pull/99'
Regression reproduces: without the fix the second merge is refused	✅ pass	live	Reverting the require_recorded_pr_identity change makes test_distinct_merged_prs_keep_distinct_wakes fail with 'distinct-merge-wakes: second merge failed'; restoring the fix makes it pass again

Evidence: fm-pr-merge PR-rebind scenario transcript (with fix + reverted-fix control)

Source: fm-pr-merge PR-rebind scenario transcript (with fix + reverted-fix control)

with fix: ok - distinct merged PRs for one task retain distinct captain-facing wakes / ok - the away record does not bypass red checks, and a recorded pr= must match the URL. fix reverted: not ok - distinct-merge-wakes: second merge failed (refused: task bound to pr-1, not pr-2)

## fm-pr-merge PR-rebind scenarios (real bin/fm-pr-merge.sh driven via run_pr_merge)

# with fix (target commit):
ok - distinct merged PRs for one task retain distinct captain-facing wakes
ok - the away record does not bypass red checks, and a recorded pr= must match the URL


# regression control - fix reverted, same distinct-merge scenario:
not ok - distinct-merge-wakes: second merge failed
(second merge refused with "error: task task-x1 is bound to <pr-1>, not <pr-2>")

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

✅ **Review** - passed

✅ No issues found.

✅ **Test** - passed

✅ No issues found.

• Live validation: ✅ go - 3 of 3 scenarios driven live against the product

Scenario	Result	Live	Evidence
After the bound PR merges, the task's next PR is accepted	✅ pass	live	test_distinct_merged_prs_keep_distinct_wakes drives real bin/fm-pr-merge.sh: merge PR/68 writes the notified marker, then merge PR/69 (pr= still bound to /68) succeeds and both leave distinct wakes
While the bound PR is unmerged, a different PR is refused	✅ pass	live	test_away_record_does_not_bypass_red_or_identity: pr= bound to /99 (no notified marker), merging /85 exits 1 with 'is bound to https://github.com/example/repo/pull/99'
Regression reproduces: without the fix the second merge is refused	✅ pass	live	Reverting the require_recorded_pr_identity change makes test_distinct_merged_prs_keep_distinct_wakes fail with 'distinct-merge-wakes: second merge failed'; restoring the fix makes it pass again

• bash runner sourcing tests/fm-pr-merge.test.sh definitions then invoking test_distinct_merged_prs_keep_distinct_wakes (drives real bin/fm-pr-merge.sh via run_pr_merge: merge PR/68, then merge PR/69 with pr= still bound to /68)
• test_away_record_does_not_bypass_red_or_identity (pr= bound to /99 unmerged, merge of /85 must refuse with 'is bound to .../99')
• Regression control: reverted the require_recorded_pr_identity fix in bin/fm-pr-merge.sh and re-ran the distinct-merge scenario, confirming the second merge is refused without the fix; restored the file afterward

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.

[greptile-apps]

Confidence Score: 5/5

[High risk] Changes how the merge script validates task rebinding.

The PR appears safe to merge.

_{Reviews (1) · Last reviewed commit: "no-mistakes(document): docs(fm-pr-merge)..."}

[kunchenguid]

Speaking as Kun's firstmate: this is merged. Thank you @karotkriss — really appreciate you taking the time on this.