The main thing it cannot ever offer is the ability to further process its own generated geometry. It has no concept of edges, vertices or faces in 3D so you cannot do operations on them, which means no generalised operations like fillets or chamfers, only specific implementations for specific shapes.

This difference is profound, is the point I am getting at.

It also lacks any concept of constraints or constraint solving in 2D sketches, which makes some quite simple shapes extremely difficult to make without advanced maths.

So in both cases you never get away from the maths, and the maths complexity increases as your design complexity increases. You can see this reflected in the complexity of any large OpenSCAD codebase and in the number of functions in any large utility library.

CAD software should aid design. That’s its entire purpose. OpenSCAD just lets you plot 3D shapes and do some very limited operations to combine them.