Seems similar in concept to the bug tedu talked about here that was exposed by OpenBSD’s randomised PIDs, where if a new thread was assigned the same PID as a thread that had just exited, the reaper would harvest the stack of the new thread.