I would guess mostly because python interpretes are just about everywhere.
Also the binaries compiled with nuitka end up being much bigger in size. A simple script of a few kb can and up in the hundreds of mb when you start compiling the dependencies, so it’s not a perfect solution.
You’re absolutely correct, and in my experience authors with physics background are even worse.
I’ve seen algorithms that I know by heart, understand fully and have implemented tens of times represented in such a way that I can’t even recognise them.