Major updates to a Linux DE has been like that forever, though.

It's easy enough to containerize an entire DE - but if you did that, you be basically running everything from inside the container - at which point you're back to square one. You're just shifting the problem from the host to the container, and the solution to fix both is the same: restore from a snapshot, reinstall, or actually try and fix the issue.

Also, a DE shouldn't bring down the whole system btw - you should always be able to switch to a second TTY to recover, and/or have a backup lightweight DE that you can switch to from your logon screen. Unless of course something really broke and caused a kernel panic and your system is fully frozen (which should be a rare occurrence on Linux-friendly hardware).

Anyways, a realistic solution would be to use an immutable distro, such as one of the Fedora Atomic/uBlue distros. The kind of breakage mentioned by OP won't be possible in such a distro, because your entire system gets updated as a single image, so it either works or it doesn't (an atomic operation), and in the event it doesn't work, you can always switch back to a previous image from the boot menu instantly. You can "pin" known good images, and this sort of image operations makes it easy to switch between latest testing/stable image version, or even switch between entire DEs with a single command. So if your KDE 6 is broken, not only can you just go back to KDE 5 with a single reboot, you can also switch to a GNOME image, or rebase to something else entirely, without messing up anything, without creating a dependency hell.