Hmm, so the failures here all seem to be: "GNOME just ignored an input".

In the install_default_update_netinst tests that failed in disk_guided_empty, openQA clicked on the 'Installation Destination' spoke in the installer, and nothing happened. The installer just stayed where it was. This is the first time the installer clicks on a spoke link in the installer hub; it did click the "I want to proceed" button on the pre-release warning earlier, though, and that worked. One interesting note here - when anaconda ported to Wayland earlier in the F42 cycle, openQA did start running into something similar mainly in the blivet custom partitioning bit of the installer, where clicks just don't do anything, and I had to add a couple of 'click again if nothing happened' workarounds. Not sure if that's related.

The desktop_browser,desktop_keyringanddesktop_update_graphical` cases are all similar. In each case, openQA switched from the desktop to a VT in order to do some prep stuff, then switched back to the desktop and hit the Super key, expecting the overview to appear. The overview did not appear. This happened twice in the desktop_keyring test, so that's four times it happened in total.

It seems like this doesn't always happen, because the browser and keyring tests passed first try on staging openQA. However, the update test failed twice. The second failure looks a bit odd and may be a timing issue in openQA, but the first failure was the same as the failures on prod. The install_default_update_netinst tests did fail the same way on stg as on prod, so that failure happened 7 times in total - two tests, each retried once on failure, across two instances (but one failure on staging actually happened earlier when anaconda got stuck in the language selection flow somehow). That one seems pretty consistent.

So I've reproduced the issues locally, they are real and reproducible, they don't only affect openQA's quirky input style. I've filed https://gitlab.gnome.org/GNOME/mutter/-/issues/3991 upstream. It seems that upstream https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4338 - which was intended to fix https://gitlab.gnome.org/GNOME/mutter/-/issues/3974 - also fixes this, but some comments on that MR (and @mclasen 's opinion) make me hesitant to just backport that PR for now. We can probably wait a few days and see what comes out of the upstream issue.