[microsoft-playwright-automation]

Avoid disabling interception from a completed handler after unrouteAll() has replaced the route list. Its own interception update must run after all removed handlers finish; otherwise the first completion sends other active requests to the network.

Extend the existing page and context regression tests to cover two concurrent requests and assert both mocked responses. All six cases fail without the fix and pass with it. Build and flint pass; the related routing suites pass on Chromium, Firefox and WebKit (325 passed, 8 existing skips).

Fixes #43878

_{Triaged on Linux by the Playwright bot - agent run}

[github-actions]

Test results for "MCP"

5194 passed, 616 skipped

---

Merge workflow run.

[github-actions]

Test results for "tests 1"

2 failed
❌ [default] › run-tests.spec.ts:1041 › should produce output twice @vscode-extension
❌ [default-trace] › run-tests.spec.ts:1041 › should produce output twice @vscode-extension

1 flaky

⚠️ [chromium-library] › library/chromium/connect-to-worker.spec.ts:19 › CR-043 should connect, evaluate, receive console and disconnect `@chromium-playwright-x64-ubuntu24-64-core-node22`

23577 passed, 704 skipped

---

Merge workflow run.

[github-actions]

🟢 Both failures predate this PR

Hi, I'm the Playwright bot. I checked the CI failures for fb60a24 against the PR diff and earlier runs.

Details

Neither failure looks caused by this PR. Both come from the same VS Code extension output assertion, not the request-interception cleanup changed here.

Pre-existing flake / infra

[default] and [default-trace] › run-tests.spec.ts:1041 › should produce output twice fail because an extra GitCommitInfo: timeout … git fetch origin … warning appears in the output. The nested test passes; the exact-output assertion rejects the warning.

The same test and warning failed before this PR's run, on unrelated commits:

• [default]: c96d4f9, run 41698335922.
• [default-trace]: 5bc53dc, run 41684939899.

The current failures have that same signature. The warning comes from the unchanged Git metadata plugin's three-second fetch limit; this PR changes only page/context route-handler cleanup and its regression tests. All six updated regression cases passed across Chromium, Firefox, and WebKit.

I refreshed and queried the aggregated test-results database. It contains no VS Code extension results, so the historical matches above are verified from the job logs. The report's Chromium worker test passed on retry and is not an outstanding failure.

_{CI failures checked by the Playwright bot - agent run}

_{Triaged by the Playwright bot - agent run}