Help needed to solve system hangs with kvm/qemu

These system hangs are starting to drive me mad, so any help would be appreciated!

I am running Debian 12 with KVM/qemu as the hypervisor in a dual GPU setup. I have a Windows 10 guest os. Windows 10 has access to a passed trough GPU (AMD 7900XTX).
When running the VM, the host pc randomly hangs. The hang is characterized by complete loss of input, displays show only the last frame and the last ~0.2s of the audio buffer keep looping. The host restarts itself after about 5 seconds.
At the time of hang (and right before it) nothing (of note) is logged to any log file I could find. Windows 10 event viewer also has nothing of note around the time of hang. Moved pstore logs also dont have anything of note.

These hangs also appear mostly random. Some days can pass and nothing happens, other times it hangs every 15 minutes.

I am fairly certain that the GPU pass-trough is the root problem, but I have no means of debugging/testing that…
Host itself (without running the Windows vm) runs fine and when I directly boot into the VM there are also no problems (VM has a whole disk assigned to it, so it can be directly booted into from the boot menu) .
Config for the GPU:

...
<hostdev mode="subsystem" type="pci" managed="yes">
  <driver name="vfio"/>
  <source>
    <address domain="0x0000" bus="0x10" slot="0x00" function="0x0"/>
  </source>
  <alias name="hostdev0"/>
  <rom file="/usr/share/vgabios/7900XTX.rom"/>
  <address type="pci" domain="0x0000" bus="0x05" slot="0x00" function="0x0"/>
</hostdev>
<hostdev mode="subsystem" type="pci" managed="yes">
  <driver name="vfio"/>
  <source>
    <address domain="0x0000" bus="0x10" slot="0x00" function="0x1"/>
  </source>
  <alias name="hostdev1"/>
  <address type="pci" domain="0x0000" bus="0x06" slot="0x00" function="0x0"/>
</hostdev>
...

The rom file is the gpu vbios dump obtained when running Windows.
I have tried to rerun the exact programs that were running when the crash occurred, but nothing happens.
At around the time of crash there is no spike in resource usage that top, htop, task manager, and other utils like that could detect.

I have kind of run out of ideas so any help would be appreciated with this!