8.x x86_64 system emulation/tcg regression (general protection fault)
Description of problem:
Running the ISO available at https://repo.chimera-linux.org/live/20230611/chimera-linux-x86_64-LIVE-20230611-gnome.iso with the above qemu command line, the graphical environment fails to come up. The system boots, and login prompt shows up; then graphical environment startup is attempted, with Wayland (you can tell as the login prompt cursor no longer blinks, being "frozen" for possibly up to a few minutes due to emulation cost). Then the graphical startup crashes (you can tell because the cursor starts blinking again) and an X11-based startup is attempted (you can tell by the X11 cross cursor) which however never fully comes up either.
Steps to reproduce:
1. Download the ISO and run with the command line above.
2. See the issue.
Additional information:
It is possible to then switch to tty2 (View->compatmonitor0, `sendkey ctrl-alt-f2`), log in as `root:chimera` or `anon:chimera` as the console prompt instructs, and type in `dmesg` (as `root`) or `doas dmesg` (as `anon`) and see that the `dmesg` contains a number of general protection faults, like this:

![Screenshot_from_2023-08-02_02-08-41](/uploads/b0e613c5191e41fce3958b74dd5dd4b7/Screenshot_from_2023-08-02_02-08-41.png)

The system used to work, but I am not sure which is the last version of QEMU where this worked, I believe 7.x. In 8.0.3 (likewise running in a Chimera environment, but it was also tested on Alpine, and I had somebody on Arch Linux test it with 8.0.2 just to rule out possible issues caused by a musl-based host environment) it crashes. It only appears to affect the `x86_64` guest architecture, as the other-architecture ISOs have graphical environment come up fine after some minutes (e.g. `ppc64le` with `qemu-system-ppc64 -M pseries-2.11,cap-htm=off -m 2048 -boot d -cdrom chimera-linux-ppc64le-LIVE-20230611-gnome.iso` works just fine). It also appears to only affect TCG emulation, as KVM likewise works fine (same command line, just `-enable-kvm` added).

Apologies for a large testcase, but it seems to need specific graphical-adjacent services to reproduce. It should be consistently reproducible though.