Earlier quoted context omitted.
I never tackled the reproducible builds topic but this is also something I had in mind while having a thought about it. Building object files for translation units, apart from macros introducing variance, should be reproducible no matter how high the number of compile jobs is running. So, isn't this largely a linking stage problem?
I would assume that compilers these days are likely to have internal parallelism too.
As someone else stated the order of compiling the object files does not matter, at least not in C or C++ compilation model. I am responsible for packaging Jank with nixpkgs, where a Clojure dialect hits the raw machine. This is done with custom LLVM implementation, but I am also working on building Jank with Zig to take advantage of its pipeline. I am being downvoted and condescended to downplaying with claims of magic while working on the frontier of PL's.