test(machine-details): fix flaky "disarms once a save settles cleanly" (PP-78tm)

[timothyfroehlich]

What

Fixes the flaky unit test MachineDetailsForm > unsaved-changes navigation guard > disarms once a save settles cleanly (bead PP-78tm).

Why it flaked

The test waited only for the "Saved" note to render, then checked that the beforeunload guard had disarmed in a separate synchronous assertion right after the waitFor:

await waitFor(() => {
  expect(screen.getByTestId("details-dirty-note")).toHaveTextContent("Saved");
});
expect(beforeUnloadWasBlocked()).toBe(false); // sampled once, immediately

The "Saved" note (machine-details-form.tsx:443-447) and the beforeunload listener teardown (:159-168) both derive from the same isDirty false-flip, but the note text lands in the commit phase while the listener teardown lands in the passive-effect phase. On a loaded CI runner the synchronous check could sample the window where the note had settled but the guard had not yet unsubscribed — a false true. It reddened the required CI Gate, including on a main commit (run 33130774346, SHA d139b903, per the bead's 2026-08-29 analysis).

The fix

Assert the disarm inside the same waitFor as the "Saved" note, so the poll retries until both conditions hold together:

await waitFor(() => {
  expect(screen.getByTestId("details-dirty-note")).toHaveTextContent("Saved");
  expect(beforeUnloadWasBlocked()).toBe(false);
});

If the guard ever genuinely fails to disarm, beforeUnloadWasBlocked() stays true, the callback keeps throwing, and waitFor times out and fails — so the test still catches a real regression; it is not weakened. The later, deliberately synchronous queryByText("Discard unsaved changes?") assertion is load-bearing and left untouched.

beforeUnloadWasBlocked() dispatches a fresh cancelable beforeunload each poll; the handler is preventDefault-only, so repeated dispatch is harmless (noted inline).

Scope

One test file, +8/-2. No component change. Comment-only rationale added; it does not over-claim the exact CI mechanism (the bead records that the split-effect cause was never reproduced locally — this is defensive hardening around the correct RTL pattern, robust regardless of the true trigger).

Validation

• pnpm exec vitest run <file>: 22/22 pass, across four reruns and once with --coverage (the CI differentiator).
• typecheck / oxlint / prettier: clean (pre-commit hook green).
• Reviewed adversarially across three lenses (correctness, bead-intent, maintainability); the one substantive finding — the comment over-committing to the unproven root cause — was folded in before this PR.

Opened unattended by the nightly bead session; carries ownerless.

🤖 Generated with Claude Code

https://claude.ai/code/session_01DeLnU9YW1LtvSN9fpE5Y41

---

Generated by Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pin-point	Ready	Preview	Sep 9, 2026 12:35pm UTC

[timothyfroehlich]

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-09T12:51:37.278294Z	76bbe44	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Can't wait for the next one!

Reviewed commit: 76bbe44f31

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".