docker-flatcar-nvidia-driver VS coreos-assembler

That leaves Flatcar and Clear Linux - both of which happen to at least have documentation for installing/running Nvidia drivers and CUDA. Flatcar has this repository from Nvidia, and I've also found this project called forklift which will supposedly handle auto-updating the kernel modules for you. The Clear Linux docs also seem to include a method to auto-rebuild the modules with kernel upgrades, though it does say that the driver version needs to be updated manually, which honestly almost sounds preferable considering how finicky Nvidia drivers can be on Linux. Clear Linux also has several other tutorials/guides that appear to try and market it for things like machine learning, which leads me to believe that Nvidia gpus would hopefully work decently on it.

If you're interested in air-gapping, I've noticed there's some innovating being done around it for kubernetes frameworks, which might include some ways to keep the machine connected without being connected, if that's your bag. Seems like a good candidate for a cattle OS though, e.g. MicroOS or Fedora Core. This is awesome:

I don't know about Flatcar but Fedora CoreOS provides good instructions for building a custom image https://github.com/coreos/coreos-assembler/blob/main/docs/building-fcos.md.

Secondly, FCOS build is documented and is easy to build your own custom base image: https://github.com/coreos/coreos-assembler/blob/main/docs/building-fcos.md. You can build anything including installers, live boot ISOs and PXE boot images.