wouldn't the name surface be better to keep?
The previous type was a pointer, no wrong structure dereferencing should cause an invalid compile due here.

wouldn't the name surface be better to keep?

It's mostly to distinguish it from the SDL surface underneath, so we avoid RenWindow.surface.surface.

The previous type was a pointer, no wrong structure dereferencing should cause an invalid compile due here.

What do you mean?

It's mostly to distinguish it from the SDL surface underneath, so we avoid RenWindow.surface.surface.
fair.

But I feel like the name rs is a bit undescriptive.

What do you mean?

Ignore it. Just Connecting dots that don't need connecting

But I feel like the name rs is a bit undescriptive.

Yeah that's fair.
What name should we use? We could even change the name of the underlying SDL_Surface.

I think it's fine; but if you must, you could simply use rensurface; that's extremely clear, and falls in line with the normal coding practice of capitalizing the class, and lower-casing the instance.

RenSurface appears to only be a combination of surface and surface_scale with a changed datatype for scales.

Does this make working with surfaces independently easier?
Does it only affect the use of the SDL Renderer?

Yes, it dissociates the surface from the window scale.
At the moment only using the SDL Renderer we get a scale that might be not 1, yeah.

I think in the future, when we refactor how we scale, we'll just not need to have a scale variable here, so then we might just replace RenSurface with SDL_Surface.

Ah, alright.

Is switching to a fixed point number going to introduce any problems?

Hopefully not.
This change was done thinking about future fractional scaling support. But considering that we might just scrap the whole surface scale concept, I should probably just revert this to be an int like before.