https://bugs.winehq.org/show_bug.cgi?id=56769
--- Comment #1 from blubban@gmail.com --- Sure is a regression-happy commit, that one. Perhaps that was not the correct solution after all.
Or perhaps we just need to power through this series of bugs. That commit didn't create any bugs, it just surfaced ones that existed already. (That's why the 5744 fixes don't revert that one, or otherwise touch the merit list.)
I'll take a look once 5744 is merged. I feel there's a high risk these fixes would touch the same places, and I'd rather not create a merge conflict with myself. (I don't want to have too many MRs open at once, either.)