It looks like a race condition in the test case. The failed test case (tests/unit/test-aio.c:test_timer_schedule()) assumes the process will be scheduled in a timely manner. A timer is scheduled for 750 ms and the test waits for 1 second before running the event loop. At that point a callback schedules the timer another 750 ms in the future. The problem is that the exact sequence of event loop (aio_poll()) invocations can be unexpected on a heavily loaded CI runner leading to spurious test failures.
I re-ran the job and it passed: https://gitlab.com/qemu-project/qemu/-/jobs/2430529496 I also ran it locally under Wine and didn't see a failure. FWIW I'm about to send another block pull request and it will include these patches again. Feel free to wait for the next pull request. Stefan