[Bugfix #1110] Fix flaky artifact-canvas D2 watch-failure test (missing waitFor)

[amrmelsayed]

Summary

Fixes a CI-load flake in the artifact-canvas test surfaces a synchronous FileAdapter.watch() failure via onError without throwing (D2).

Root Cause

The test awaited onError firing, then asserted the DOM synchronously on the next line:

await waitFor(() => expect(onError).toHaveBeenCalled());
expect(document.querySelector('p[data-line]')).not.toBeNull(); // bare read — race

ArtifactCanvas's async read() render can resolve after onError fires. Under CI load the render is delayed enough that the synchronous DOM query runs before the paragraph is in the tree, so the assertion intermittently fails. The sibling test directly below (Disposable.dispose is safe to call more than once) already uses the correct waitFor-wrapped pattern.

Fix

Wrap the second assertion in waitFor to match the sibling test's pattern:

await waitFor(() => expect(onError).toHaveBeenCalled());
await waitFor(() => expect(document.querySelector('p[data-line]')).not.toBeNull());

Semantics are unchanged (the assertion still must eventually hold); only the timing tolerance widens. One-line change, no production code touched.

Test Plan

• pnpm --filter @cluesmith/codev-artifact-canvas test → 56/56 pass.
• Stress-ran the D2 test in isolation 5× → 5/5 pass.
• Full workspace pnpm build + pnpm test green (verified via porch checks: build ✓, tests ✓).

Regression Coverage

This is a test-timing fix to a race, so the corrected D2 test is the regression coverage: with the bare synchronous read it flakes under load; with the waitFor wrapper it is deterministic. An additional standalone test would only re-assert the same behavior.

Fixes #1110

[amrmelsayed]

CMAP review (3-way)

Model	Verdict
Gemini (agy)	APPROVE
Codex	APPROVE
Claude	APPROVE

All three approved the one-line fix. Claude's review flagged the identical bare-assertion-after-waitFor race in the sibling test surfaces adapter errors via onError and does not throw; failed list keeps prior markers (D2) (line 133). The issue explicitly invited fixing such patterns as a ride-along ('the file is small'), so I applied the same one-line waitFor wrap there too (commit 846b53d). 56/56 tests pass.

[amrmelsayed]

Architect Review

Low-risk BUGFIX. Two-line diff in packages/artifact-canvas/src/components/__tests__/artifact-canvas.test.tsx, exactly as scoped:

• Line 161 — the D2 watch-failure test originally reported. expect(document.querySelector(...)).not.toBeNull() is now wrapped in waitFor(() => ...).
• Line 130 — the sibling D2 adapter-error test with the identical race. Same wrap. This is the ride-along the issue's Out-of-scope section explicitly invited ("could ride along if the fixer notices any") — picked up via the CMAP claude lane.

Three-way CMAP unanimous APPROVE (gemini + codex + claude). 56/56 tests pass. No production code changes; semantics preserved (the assertion still eventually succeeds, just with retry tolerance matching the sibling test's pattern).

APPROVE. Ready for the pr gate.

---

Architect review