[TRIAGE TEST 18870] tests: thread/stress_heap.py intermittent failure on macOS

[andrewleech]

The stress_heap.py test was excluded from the macOS CI job with a "is
flaky" comment prior to the analysis period. No direct log observations
are available since it was already excluded, but it's listed in
FLAKY_TESTS restricted to the darwin platform.

PR #​18861 now handles this via ignore-on-failure instead of exclusion,
so the test runs and its output is visible but doesn't block CI.

See analysis: https://gist.github.com/andrewleech/5686ed5242e0948d8679c432579e002e

---

Cloned from micropython/micropython issue 18870 for triage testing.

[andrewleech]

/triage

[mpy-reviewer]

Labelled bug / tests / port-unix (macOS is the unix port).

PR micropython#18861 mitigates CI noise by switching from exclusion to ignore-on-failure, so the test now runs and produces visible output. That's a workaround — the underlying failure is unresolved.

Root cause is unknown: thread/stress_heap.py stresses the GC under multi-threading on macOS. The failure mode and frequency aren't documented here. A few things would help narrow it down:

• What does the actual failure output look like (assertion error, crash, hang/timeout)?
• Is the failure rate higher under the macOS 15 runner migration (post-Aug 2025)?
• Does it reproduce locally on macOS with MICROPY_CPYTHON_COMPAT=0 ./run-tests.py --jobs 1 thread/stress_heap.py?

Related open issues in this series: micropython#18866 (thread_gc1.py race), micropython#18867 (QEMU stress tests).