I'm not a game developer, just a regular application developer, but wouldn't an overlay system that works like OOP fix the variation issues? I suppose that instead of drawing each scene individually, you could create body parts individually and just customize those for each unique scene. The colors would just get mapped to each part of the body.
I guess the problem is just as you said, permutations. Since it is 2D art, as opposed to 3D modelling. Also, I wonder if the Live2D (not exactly using that software) strategy would work. Create a sort of 3D model first, then let the system create the 2D variations as assets.
Don't worry about this, it's just my brain trying to think about how to help with software engineering side of games. Great game though! 🥰