It's not a big problem in Python in general, only in scientific computing / number crunching projects, because of the dependencies on huge complex software, som

It's not a big problem in Python in general, only in scientific computing / number crunching projects, because of the dependencies on huge complex software, some of it ancient, written in C, Fortran, and C++. So why do we hear about this problem in Python a lot? Well, because it's what's used for the glue/frontend, which is what users work with directly. It's selection bias. Sure, another language might fare somewhat better or worse for this or that reason, but at the end of the day it's gonna be a pain in the ass (at least until next-generation, complete, deterministic, language-agnostic solutions like Nix/GUIX really gain traction).