It sounds like they did significant work getting the whole stack to work well, from Wayland to the RDP back-end to improvements to the RDP client on the Win32 side.

Your interns must be rock stars on meth.

Most likely energy drinks + Adderall + financial insecurity. Adderall, Meth, same thing.

That sounds like the killer app. GPU passthru can be an absolutely gnarly undertaking with any virtualization system.

GPU passthrough on WSL2 is a thing since around December 2020 -- but I'm not sure if it's in stable yet, I had to install an insider build for it. It works surprisingly well and I was able to develop my ML project using it (with the help of VSCode devcontainers)

Why did you want to develop it through WSL2 instead of just compiling it on windows?

Anyone could already do it with a few lines of code.

Edit: Seems people don't believe me, here's an article how to do it: https://techcommunity.microsoft.com/t5/windows-dev-appconsul...

That shows how to set up an X server in windows that WSL can access. How do you get GPU acceleration for ML tasks using X11?