Bind subdivision surfaces to skeletons
For optimal performance, bind subdivision surfaces at a coarser level, and only to one level. This usually results in adequate shape changes as the deformers and joints move, and gives you fewer vertices to weight while maintaining smooth interpolation.
The detail edits on finer levels tend to float or ride on the bound coarser level, which helps skin details such as wrinkles and spots blend smoothly. For example, as a finger joint bends, the wrinkles riding on top, modeled at a finer level, stretch out.