Some tests seem to set up a failloc race condition and then reset it before the other thread has a chance at being woken up therefore hanging the thread and causing all sorts of cleanup failures. e.g. sanityn.sh test 35 is an example of that (LU-13328)