Something spurred a thought recently (I seem to recall a siggraph announcement about some software package having animatable uvs), that if the uv coordinates could be weighted and attached to the skeletal system in such a way that they only partially follow the actual surface deformations, that it could give a much more realistic appearance closer to the actual simulation of skin with a muscular structure.
Maybe the uvs should be weighted to their corresponding surface instead (this might require a different kind of weighting: ie. rest versus change, so that under normal circumstances the texture is in the correct position, but as the surface shifts it partially follows).
Anyways, just a thought. Would this be feasible/practical/a good idea?