https://bugs.winehq.org/show_bug.cgi?id=53020
--- Comment #5 from Rafał Mużyło galtgendo@o2.pl --- While it seems that particular commit shouldn't have any effect, as calls to wait_message have already been removed, there's a chance, that the tricky part is nulldrv_MsgWaitForMultipleObjectsEx.
The condition checked for there for NtYieldExecution is 'ret == WAIT_TIMEOUT && !count && !timeout'. In nulldrv_MsgWaitForMultipleObjectsEx of that commit, that was an early return without going into actual MsgWaitForMultipleObjectsEx.
However, in nulldrv_MsgWaitForMultipleObjectsEx of win32u/driver.c, the condition for the early return is '!count && timeout && !timeout->QuadPart'... Then again, here MsgWaitForMultipleObjectsEx is called with '!(flags & MWMO_WAITALL)' whereas before is was without negation...