FWIW, here is the upstream merge request fixing the reported regression: https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/4152

Gnome remote desktop 47.2 was just pushed, does it still happen with that one as well?

It's a race condition in the frame scheduling routines in the compositor, unrelated to gnome-remote-desktop. It just happens to (in bojan's case) hit the race more easily. Didn't ever run into it here, and had to craft a special purpose test case to replicate it.

Yes, https://gitlab.gnome.org/GNOME/mutter/-/issues/3804, I discovered a work around that Bojan verified working, and am currently working on a real fix.

Are you connecting via gnome-remote-desktop running in the VM?

@bojan, could you expand what you mean not working? Running with this update, I can still launch a headless session and connect to it via RDP.

The helper to run the whole session was decide not to be part of the gnome-remote-desktop package itself, but gdm, so that it could be generic. I suppose we need to add that back to the package downstream if it didn't end up in gdm. Can you open a bugzilla issue about it?

Update fixes using this input method for me.

This update has been unpushed.

This update has been unpushed.

There is no executable in /usr/bin that needs a manual; gnome-remote-desktop-daemon is an executable only used by a systemd service.

It was not cherry picked, as it is not ready yet. However, a gtk update will resolve many of the issues that merge request tries to address. Merge request 801 will hopefully be backported later when it has been made sure not to regress.

Should be replaced with https://bodhi.fedoraproject.org/updates/FEDORA-2019-4c65e7710a

Oh, I couldn't..

Intending to replace this update with one that contains a fix to a bug introduced by it, so makring it with negative karma.

Update fixes #1641989 for me.