https://bugs.winehq.org/show_bug.cgi?id=33479
--- Comment #37 from Henri Verbeet hverbeet@gmail.com --- (In reply to Sebastian Lackner from comment #35)
I suspect that this problem occurs when a WM_X11DRV_CLIP_CURSOR message is interpreted incorrect. Does this patch help?
For what it's worth, I'd expect the issue to essentially be that ClipCursor(NULL) is used both for fullscreen clipping by applications and for ungrabbing the clipping window on process exit. (The latter happens from the server as "set_clip_rectangle( desktop, NULL );".)
That could also be why for at least some applications having a dualhead setup might make a difference as claimed by comment #33, since in that case the application is likely to pass something smaller than "virtual_screen_rect" to ClipCursor().
If that's the case, the way to fix it would probably be to signal ungrabbing the clipping window in some other way.