Microsoft seems to be taking a outside-in "component at a time" approach to open sourcing Windows. Terminal, Notepad, Paint, Calculator, the new Edit.com replacement, a lot of WSL now, etc.
This approach has been fascinating so far, but yeah not "exciting" from "what crazy things can I do with Windows like put it in a toaster" side of things.
It would be great to see at least a little bit more "middle-out" from Windows Open Source efforts. A minimal build of the NT Kernel and some core Windows components has been "free as in beer" for a while for hobby projects with small screens if you really want to try a very minimal "toaster build" (there's some interesting RPi out there), but the path to commercialization is rough after that point and the "small screens" thing a bit of a weird line in the sand (though understandable given Microsoft's position of power on the desktop and sort of the tablet but not phone).
The NT Kernel is one of the most interesting microkernels left in active use [0], especially given how many processor architectures it has supported over decades and how many it still supports (even the ones that Windows isn't very commercially successful on today). It could be a wealth of power to research and academia if it were open source, even if Microsoft didn't open source any of the Windows Subsystems. It would be academically interesting to see what sort of cool/weird/strange Subsystems people would build if NT were open source. I suppose Microsoft still fears it would be commercially interesting, too.
[0] Some offense, I suppose to XNU here. Apple's kernel is often called a microkernel for its roots from the Mach kernel, but it has rebuilt some monoliths on top of that over the years (Wikipedia more kindly calls it a "hybrid kernel"), and Mach itself is still so Unix flavored. NT's "object oriented" approach is rather unique today, with its more VMS heritage, a deeply alternate path from POSIX/Unix/Linux(/BSD).