Ok, so did some more testing today:
Official Plex Docker
thor2002ro 5.18 22020526 Kernel
Unraid 6.10.0
Intel_gpu_top installed AND still had modprobe appended in syslinux config.
The above config would crash plex and eventually unraid.
Linuxserver docker
thor2002ro 5.18 22020526 Kernel
Unraid 6.10.0
Intel_gpu_top installed AND still had modprobe appended in syslinux config.
The above config would crash plex. I didn't try long enough to see if unraid crashed as well.
Linuxserver docker
thor2002ro 5.18 22020526 Kernel
Unraid 6.10.0
Intel_gpu_top installed AND REMOVED modprobe in syslinux config.
After removing modprobe from syslinux config plex now transcodes without crashing - only tested a few files that previously crashed plex.
So, I don't know if it's the change to the docker container or that I had the modprobe statement still in the config that was causing problems. It's also possible that this is just a fluke and it'll stop working soon.
One thing I do notice, when it isn't working correctly: if I watch the GPU statistics plugin, there will be some render load and video load but then the usage will stop, however the power will stay at around 2W. At this point, any attempt to transcode will crash plex. The GPU is also locked at 300 MHz (viewed using intel_gpu_top). So it's as if the GPU is getting stuck and is no longer accessible by anything else.
When it's working correctly, there will be render load and video load with power up to 4W. This will turn on and off as it reaches the buffer limit. But the power always goes to zero when it's no longer transcoding.