Comment 13 for bug 820327

One of those traces - with-displays-applet-gnome-settings-daemon.trace - is with the displays capplet open, right? And that one works correctly?

It seems that g-s-d is submitting an identical set of RANDR requests; the differences are in the other events that occur. I suspect that this might have the same root cause as bug 828623 and bug 861426 and that this seems to be a timing-related issue in the driver or kernel (probably the kernel) around modesetting.