Fix flaky test_siem_export race condition
`apps/api/tests/test_siem_export.py::test_polling_the_export_writes_no_read_row_into_the_org` failed once in a full `./dc test -n 9` run on develop e7ceb65 (`assert 0 == 1` on `captured_writes.rows("page_view", path=control)`), and passed 36/36 when its file ran alone. The control read's page_view row probably lands off-thread after the assertion. Make the test wait for that write deterministically, or capture it synchronously, so it cannot race.
Out of scope: changing the SIEM export or the user_activity fence.
Done means: the test passes in 5 consecutive full `./dc test -n 9` runs (or in a targeted stress run that reproduced the race before the fix); `./dc check quick` green.
Outcome
Fixed the test_siem_export race: the control read capture was cancelled when the per-request portal loop closed, so the test now captures it inline. Stress repro went from 2/200 failing to 0/400; quick gate green. PR #7850 queued to the merge lane.
Artifacts
Activity
- Created
- Run
- Runner started
- Runner ended (done)