We had a timeout in the test StressAsyncUL yesterday on Linux aarch64 and one on AIX a few days ago.
Timeout happened always in the 4th part of the test (analyze_output(true, "-Xlog:async:stall", "-Xlog:all=trace", "-XX:AsyncLogBufferSize=192", InnerClass.class.getName());) .
Probably we should increase the whole test timeout a bit so that it is stable too on slow machines.