qa: `tool_bench_sanity_check.py --bench=MempoolCheck` failure #34248

On NetBSD 10.1, https://github.com/hebasto/bitcoin-core-nightly/actions/runs/20872235098/job/59975888965:

90/455 - tool_bench_sanity_check.py --bench=MempoolCheck failed, Duration: 26 s

stdout:
2026-01-10T04:19:37.691064Z TestFramework (INFO): PRNG seed is: 4615966168349338210
2026-01-10T04:19:37.692250Z TestFramework (INFO): Initializing test directory /home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/test_runner_₿_🏃_20260110_041921/tool_bench_sanity_check_363
2026-01-10T04:19:37.692513Z TestFramework (INFO): Starting: /home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/build/bin/bench_bitcoin -filter=MempoolCheck -sanity-check
Running with -sanity-check option, output is being suppressed as benchmark results will be useless.
2026-01-10T04:19:38.567933Z TestFramework (INFO): Success!
2026-01-10T04:20:03.637759Z TestFramework.utils (ERROR): wait_until() failed. Predicate: ''''
        wait_until_helper_internal(lambda: not self.network_event_loop.is_running(), timeout=timeout)
'''


stderr:
Traceback (most recent call last):
  File "/home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/build/test/functional/tool_bench_sanity_check.py", line 41, in <module>
    BenchSanityCheck(__file__).main()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/test/functional/test_framework/test_framework.py", line 154, in main
    exit_code = self.shutdown()
  File "/home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/test/functional/test_framework/test_framework.py", line 281, in shutdown
    self.network_thread.close()
    ~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File "/home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/test/functional/test_framework/p2p.py", line 749, in close
    wait_until_helper_internal(lambda: not self.network_event_loop.is_running(), timeout=timeout)
    ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/bitcoin-core-nightly/bitcoin-core-nightly/test/functional/test_framework/util.py", line 420, in wait_until_helper_internal
    raise AssertionError("Predicate {} not true after {} seconds".format(predicate_source, timeout))
AssertionError: Predicate ''''
        wait_until_helper_internal(lambda: not self.network_event_loop.is_running(), timeout=timeout)
''' not true after 10.0 seconds

Not sure how this could happen. After https://github.com/bitcoin/bitcoin/commit/39ca01525977e5aab078198874f6bf035b552b3b , the test waits for the loop to be running, so a stop should always be called on a running loop and result in it eventually not running.

@maflcko, I think yes, after https://github.com/bitcoin/bitcoin/commit/39ca01525977e5aab078198874f6bf035b552b3b a stop() will always be called on a running loop. I think what is happening here is:

self.network_event_loop.call_soon_threadsafe(self.network_event_loop.stop)
wait_until_helper_internal(lambda: not self.network_event_loop.is_running(), timeout=timeout)

the second line times out, "not true after 10.0 seconds". So it just did not stop in 10 seconds? In other words, that "soon" in call_soon_threadsafe() is not soon enough?

Heh, the simplest cause could be the correct one. A 10 second delay due to resource contention seems a bit high, but this is running a VM (bsd) inside another VM (GHA), so it seems plausible.

I guess scaling here with the timeout factor may be correct anyway. Done in #34369

If this happens the next time in the nightly repo, the timeout should be 80 seconds here. If that happens, we can start looking again.