Speed increases way too fast when you go smaller. That's the only problem. It would probably work with a very open layout with far away ramps and nothing too close by, but otherwise it doesn't lend to more interesting design choices. I think the current size really allows the optimal footprint vs speed. Larger, and there would be no way to move it around (Hercules).
I actually tried messing around with 3/4 size bearings on a piece of MDF with some flipper assemblies attached while entertaining the idea of an arduino based pitch and batt. Needless to say, it didn't work well at that size, and I abandoned it shortly after that initial test, since the entire PF was built around the concept of that size ball.
But I hear you on sameness. No real innovation going on. But that is what people like and what sells. We won't see further innovation in pinball PF layouts/size probably because these aren't 99% in arcades generating large profits anymore where innovation could put a leg up on the competition for quarters.