BTW, what is the difference between the phase 5 and the commodore 68040 libraries, what do they have to do differenty to work?
P5 680x0.library has some new API functions to allow the P5 commandline tools to modify MMU mapping in a consistent way. Also, the P5 libs include support for more hardware (obviously, since they're newer). Also, the 0x0 libs in BPPC also use some special MMU map setup to allow system to work properly with the dual 68k / PPC setup.
How is the BPPC's 040 different to the A4000/040 or a standard 040 accelerator card?
From 040 side of things, nothing much. MMU setup is a bit different to most other cards, but nothing user programs should worry about.