[Nir-Az]

Summary

unit-tests/live/metadata/pytest-sync.py::test_synchronized_frames segfaulted in nightly LRS_linux_compile_lib_ci #11175, killing the whole pytest session. The Python main thread was sleeping in time.sleep(5) (line 161) immediately after device.hardware_reset() (line 160); the actual fault is in a librealsense backend thread.

The pytest hub-recycle in module_device_setup (unit-tests/conftest.py:435-466) had just power-cycled the port and re-enumerated the D455 milliseconds earlier — calling hardware_reset() on a freshly enumerated USB device triggers a race in the SDK backend.

Changes

• Drop the per-iteration device.hardware_reset() from the loop in test_synchronized_frames — it was inherited verbatim from legacy test-sync.py, which predated the hub-recycle infrastructure. With pytest's --hub-reset, iter 0 already gets a freshly recycled device, and pipeline.start/stop isolates per-config state for iters 1-5.
• Add time.sleep(1) before the loop to let the device settle after the hub recycle (matches the convention in pytest-stress.py:75 and pytest-sanity.py:48, both of which pass).
• Add time.sleep(2) between iterations to let hardware settle after pipeline.stop() before the next configuration starts.
• Kept the existing time.sleep(5) after pipeline.start() in run_test() unchanged.

Reference

• Jira: RSDSO-21449 — SDK-side race between udev device-removed callback and in-flight USB transfer; tracked separately. This PR is the test-side mitigation; the SDK fix is a separate workstream.

Test plan

• Wait for libci CI to run; confirm pytest-sync.py no longer segfaults under --context nightly --hub-reset.
• Confirm the test still exercises all six (resolution, fps) configurations.

Nightly build passing - https://rsjenkins.realsenseai.com/job/LRS_libci_pipeline/15315/

[Copilot]

Pull request overview

This PR mitigates a reported nightly segfault in the pytest-sync live metadata test by removing a redundant per-iteration hardware_reset() that can race with the hub-recycle infrastructure, and by adding small stabilization delays around pipeline reconfiguration.

Changes:

• Remove device.hardware_reset() and device re-enumeration from each configuration iteration in test_synchronized_frames.
• Add an initial time.sleep(1) after fixture setup to allow post hub-recycle settling before the first pipeline.start().
• Add time.sleep(2) between configurations to allow settling after pipeline.stop().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[AviaAv]

LGTM