Multi monitor setups have always been a problem on both Windows and Linux. Always.
Oftentimes you will get lucky and the right combination of drivers and hardware combine to create a stable experience for you. But there are no guarantees for that and if it works, it’s by sheer dumb luck. I can name plenty of other times when windows jump around on Xorg and Windows as well. Especially with laptops.
The only stable way for windows to stick in the same position consistently (e.g. works for every screen setup) across monitors, sleep mode and unplugging / replugging the laptop dock I have found is to go the route of i3; Every virtual desktop fits on exactly one screen, and then the windows are positioned in that virtual desktop and it is the virtual desktop that get assigned to separate screens.
This and an improved way to do native tiling has been on my wishlist card for Gnome since 2015 or so. I hear that steps are finally being taken to fix this in both KDE and Gnome, which is great news.
However, whether you like it or not, Xorg is going away. What you are seeing now are the last transition hurdles as apps that refused to touch wayland are finally forced to take the step.
We saw the same issues with pure ALSA → PulseAudio, and a lot of those issues were because ALSA was a godawful hack and spagetthi code everywhere that made it almost impossible to implement something better on top of it. The PulseAudio devs persevered, but late 00s was not a fun time to run Ubuntu as sound was pretty borked as apps slowly switched over, emulation layers were perfected and the new way of things made its way into most distros.
wayland is both in kind of the same situation, and not. Xorg is now being phased out because the pain of fixing Xorg is greater than the pain of adapting to wayland. That does not mean wayland is perfect. Some is from lack of support or an incompatible way to do the same thing, and some of the friction is due to differences in philosophy, like the decision that an application should never care about window placement - and if it does, that is an outdated concept and now considered legacy behavior.
As with any new rule, you might think it is stupid and as always there will be a gazillion ways to break any rule - but it is the rule because the devs made it that way. Most rules in wayland are not arbitrary, but chosen with a focus on privacy and security.
And btw, Radeon is the gold standard here with regards to desktop and ecosystem support. Intel is trying but doesn’t quite have the resources. Nvidia is dragging their feet as always, their solution is currently “working but…”
So the TLDR; - Yes, not perfect. But getting there, and hopefully certain GPU and software vendors / projects will remove that stick in their bum, roll up their sleeves and get to work creating a decent Linux environment again.
However, my example would take about 30 seconds to test and only involves doing the one thing that I declare is more painful than it’s supposed to be… Meaning them being silent about how testing went is a clear data-point indicating that they know testing will not go well for them.
