test: repair remote and update-check fixture assumptions

[Ikha14]

Summary

Repair three independent test-fixture assumptions exposed by full-suite run 01M0NK3HAGHN2HB06QXHCQN75H:

• copy the optional fm-ps-lib.sh dependency into the remote handoff fixture when a downstream port provides it, preventing the copied wake library from failing before the confined-put assertion;
• prove the orphan-worker scenario by recording and observing exit of the actual launch shell instead of requiring Linux to reparent the worker specifically to PID 1, because a process subreaper may validly adopt it;
• give the update-announcement budget fixture one full clock second for setup before its stalled version probe consumes the budget, avoiding a whole-second boundary race that could stop before the watched source was reached.

The documentation failure was not a canonical documentation defect: the failing workspace contained unrelated .paul/*.md files from a divergent downstream default branch, while the tested branch and canonical inventory contain none of those files. No documentation text or inventory assertion was loosened.

Diagnosis

• Remote handoff trigger: the mixed workspace supplied a wake library that sources fm-ps-lib.sh; masking condition: the fixture copied only the older dependency set; symptom: the superseded-payload comparison ran after fixture initialization had already failed.
• Orphan reap trigger: a detached worker is launched; masking condition: the full-suite process tree uses a subreaper; symptom: the fixture rejected a genuine detached worker solely because its parent was not PID 1.
• Tool update trigger: a one-second sweep starts near a whole-second rollover; masking condition: registry/setup overhead consumes the fractional remainder; symptom: the sweep reports it never reached NoMistakes rather than the more specific unqueried-announcement source.

Validation

All commands were bounded by timeout 600s.

• bash tests/fm-remote-backlog-handoff.test.sh - passed
• bash tests/fm-remote-job-orphan-reap.test.sh - passed
• bash tests/fm-documentation-audiences.test.sh - passed
• bash tests/fm-tool-update-check.test.sh - passed, including three consecutive supporting runs during diagnosis
• bin/fm-lint.sh - ShellCheck 0.11.0 and actionlint 1.7.12 passed
• bin/fm-doc-audience-check.sh - passed with 73 surfaces and 266 local links
• git diff --check upstream/main...HEAD - passed

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with no concrete changed-code defect identified.

The changes are confined to test fixtures and preserve their intended assertions while accommodating optional downstream dependencies, process subreapers, and whole-second budget boundaries.

_{Reviews (1): Last reviewed commit: "test(tool-update): stabilize exhausted a..." | Re-trigger Greptile}