Several TreeSE methods appear to be slow. We should identify the ones that can be accelerated best, and then optimize their speed.