They can scale, but not every code base is well written. Operating from the assumption that best practices is always implemented isn't necessarily true. Sometimes you work with sub-par code or code that was written by someone skilled but due to time pressure; misunderstanding the requirements; or a poorly planned architecture means that it does not lend itself to scalability. That's the problem. You can refactor it, but refactoring code in Python can be a nightmare compared with tools like Visual Studio that make it a breeze -- plus you get the benefit of having a compiler catch errors for you.

Well-thought-out code will always trump bad code; but that's a truism.