It occurs to me that a really easy solution to Oxide’s actual futurelock bug (which used a mpsc sender rather than a mutex) would be to have the channel wake up every waiting sender instead of only the first. Not sure how much worse it would be in a non buggy case.