test: Fix flaky test_lifecycle_on_platform_without_websocket

[vdusek]

test_lifecycle_on_platform_without_websocket pointed the events WebSocket URL at a hard-coded ws://localhost:56565 and asserted that the connection is refused. That port sits inside the OS ephemeral range (32768-60999 on Linux, 49152-65535 on Windows), so any bind(host, 0) elsewhere in the parallel suite can be handed exactly that port, and this test file alone starts three such servers. When a sibling xdist worker got 56565, the connection succeeded and the test failed with DID NOT RAISE RuntimeError, as in this Windows run. Attempt 2 of the same run passed on the same commit.

The test now reserves a port instead of guessing one. A new _unreachable_ws_url helper binds ('127.0.0.1', 0) and holds the socket open without calling listen(). Holding it keeps the OS from handing the port to anyone else, and a bound socket that never listens refuses every connect with ECONNREFUSED, which is the OSError the test asserts as the cause. The assertions themselves are unchanged.

Running a real WebSocket server on 127.0.0.1:56565 reproduces the CI failure exactly, which gives a before/after: 20/20 runs failed before the fix, 0/100 after. The fixed test also survived 0/25 whole-file runs under --numprocesses=auto with that port occupied, and 0/60 runs with a background process holding ~400 rotating ephemeral listeners.

✍️ Drafted by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.78%. Comparing base (8532389) to head (d43fe2d).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1121      +/-   ##
==========================================
+ Coverage   92.75%   92.78%   +0.02%     
==========================================
  Files          53       53              
  Lines        3519     3519              
==========================================
+ Hits         3264     3265       +1     
+ Misses        255      254       -1     

Flag	Coverage Δ
e2e	32.73% <ø> (ø)
integration	53.36% <ø> (+0.02%)	⬆️
unit	89.34% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.