Rust or Haskell. Python is subpar in terms of maintainability especially when it comes to long-running, generational, massive, codebases. Python is also massively worse when it comes to performance. Haskell strikes the perfect balance here, I've worked at companies with massive codebases where maintaining is still a breeze years later. This simply does not happen with any OOP languages, the code always turns into an ingrown hedge with time.

Especially when it comes to long-running projects like what's attempted here, Python, even being a technology from yesteryear, isn't even suited for problems of yesteryear.

With projects like these maintainability is your #1 concern, so referential transparency (greatly reinforced by laziness-by-default) and purity are indispensable and irreplaceable here.

The next concern is adequate performance, and Haskell has that in heaps.

The final concern is the capability to find programmers, and there are heaps of those in Haskell land.

Well I still think it's something to worry about before you end in front of a "Your account has been suspended" notice. How can you be sure someone has the whole dataset locally backed up ? Plus, you still have to reconfigure the workflows and stuff like that. Not to mention that sometime, project's websites are also hosted on github, so that's something else you have to rebuild.

And also, this means devs/contributors have to give their data to MS.

Since the whole point of the project is sovereignty, it should have started with setting up their own gitea/gitlab/etc. instances and domains.