In the last testruns I got the following tests failing but passing on rerun (sometimes the n-th time):

eventfd
async_signal_syscalls2-32
check_lost_interrupts
thread_stress
thread_stress-32
checkpoint_mmap_shared
condvar_stress-no-syscallbuf
mmap_adjacent_to_rr_usage
mutex_pi_stress
simple_threads_stress-32

I've got similar issues (also with other tests) before on several machines with different CPUs.
What I've recognized is that this is much more likely to happen when I run ctest with multiple jobs (more jobs, more likely).

Question: Should those failures just be ignored (and the wiki adjusted to always rerun the failed tests with a single job)?
If not: how to analyze this best?