Comment 22 for bug 967416

Do we have a plan for fixing this? Threading bugs can be so tricky...

Would it make sense to have the save routine set a lock when it starts, and then have the exit thread check for existing locks and refuse or delay exiting until they've cleared? Perhaps the lock could be timestamped so that the lock is only honored for, say, 10 seconds (the save routine could re-issue the lock if it expects the save operation to take more time than that.