When an asynchronous event can't be queued for dispatch because the queue is full, stack traces are logged for all of the event threads to show what they were doing. It appears there are some subtle race conditions in that handling which can lead to some threads not being logged at all, and others being logged multiple times.

Additionally, since event handlers can raise other events, it's possible for the rejection that triggers logging to occur on an event thread, in which case that thread is delayed from handling other events until it finishes logging what the other event threads are doing.