For AGA speedup, i suggest you BlazeWCP, FBlit and Ftext.
These are the three patches that I use, and other than MCP (definitely NOT a performance tool!), I don't have any other patches on my systems. I definitely recommend that you at least use these three!
I run a 256-colour workbench with 256-colour backdrop, with solid window moving and resizing, and it's nippy, even without the patches. Both my systems (A1200 and CD32) are expanded to 50Mhz 030's with 50Mhz FPU's and 32MB RAM (ie. more or less identical) and yet my CD32 can run a 256-colour workbench about two or three times faster than the A1200 can... the Akiko chip, maybe? or perhaps something related to the the SX32 Pro?