The test heavily relies on timers (e.g., curTimeMicros64()) to test the order of events. This could potentially cause failures if the timer does not advance between two events (e.g., see here and here).
This ticket should improve the usage of timers in this test to make sure a non-increasing timer would not cause test-failures. A potential solution is to introduce artificial delays between the points of interest, as previously done by
This test is only available on 3.6, 4.0, 4.2, and 4.4.