This should now pass the tests, as shown here https://testbot.winehq.org/JobDetails.pl?Key=120730. The issue was that the `winex11` driver extra data layout isn't portable and so reading it from the registry doesn't work. I've kept the `winex11` `get_full_mode` helper call for now.
For XRandR I believe RROutput and RRMode are XID and should be valid across process (though arguably if the X11 display isn't the same, it may not work), but for XF86VM there's possibly some private data in XF86VidModeModeInfo which I'm not sure can be.
Maybe the idea of putting the driver data in the registry is just wrong.