This has given me a lot to think about. Time and money aren't really limiting factors at the moment (my wife is being very patient :P), so my only major concern is doing things correctly. With that said, I'm actually working on two arrays concurrently. The other is a 7x21 matrix of red LEDs, and I think the multiplexing method would work much better for those, since it's basically just a glorified scrolling marquee.
I think for the RGB array I'll stick to individually addressing anodes, since both brightness and color resolution are going to be significant factors. The animations will also be much more complex, and reducing software complexity a little will save some time. I'm a software developer by trade, so I'm sure I can manage enough abstraction to make things develop somewhat smoothly.
Another issue I didn't think was worth mentioning was that these displays will actually be mounted to two helmet visors, so being able to see through them (and thus, reducing the number of wires and components) is a consideration.
Power is another issue I had been avoiding until I had a finalized schematic in mind, but I suppose I should have been considering it. I don't expect that 5v will probably be enough in the end, but I really haven't even scratched the surface of that.