This issue is similar to SERVER-120972, but in another test case: ComplexPermitForwardingWithMixedOutcomes, the intend of the test is more focused towards unordered_ticket_semaphore, because the release only wakes up threads, in the case of ordered_ticket_semaphore the class is designed to ensure acquisition order, so, doing a release in some machines might race the timeout of the test. It is important to notice that this might also happen in unordered_ticket_semaphore but it's significant less likely.

At this point I think we should simply remove that test too.