When LEWG looked at issue 3417 in Brno, it was suggested that instead of just deprecating (and then removing via P2866) non-lock-free wait/notify on volatile atomics, we should remove all wait/notify on volatile atomics. If the value of a volatile atomic object changes spontaneously (as volatile objects are wont to do) then threads blocked waiting for it to change will not be notified. This makes using wait/notify on volatile objects unreliable and of questionable value.