one way I optimised my own ProTracker playroutine was to take out all support for fine tuning. Those period tables were just taking up too much space! And I always make a point of tuning my samples properly anyway.
What I always wanted to do was use the Copper to handle the repeat pointers, but I never got it to work. DMA wait loops are so ugly. I might just rewrite the whole thing someday but it's not a high priority.