[wizeman]

That makes a lot of sense, thanks for all the answers!

It's definitely a very interesting project and I think that once more programs can work without changes, it is something that could perhaps be used for building packages by distros that have displayed a special interest in (and can benefit significantly from) reproducible builds -- like Debian for instance, but especially NixOS due to the existing work on content-addressed paths (explained here: https://www.tweag.io/blog/2020-09-10-nix-cas/ ).

The serialization of threads (and I assume processes?) would probably be an issue for large packages, but at least it would be possible to build many packages in parallel once the base dependencies have been built.

I had thought about starting such a project myself (for reproducible builds), but of course, the amount of work is very significant and for me it was not worth the benefits :)