https://bugs.winehq.org/show_bug.cgi?id=56394
--- Comment #3 from Chiitoo chiitoo@gentoo.org --- A bit of a follow-up on this...
It seems that the issue is still there, /sometimes/, and I can now see it happening even with the parent commit of the SHA currently marked as the regression introducer.
There must be some unknown factor to it all, since when testing this, I could not make it happen while trying for several minutes, several times, and then right away with the mentioned change applied.
Will try to dig more into this, as time and energy permits.