I guess I can see it.. one lock would guard invalidation<->fork and the other would guard fork<->crash (and not necessarily need to do the same for invalidation, I suppose?)