Seems like they should have used a compute module instead of building out a new PCB? Any idea why that wasn't done? Pretty damning not to dogfood your own product.
I don't think it's a dumb question, though I guess some might question its relevance. Side-comments can be interesting, though.
I don't know why they did it, but it follows the same pattern as the pi400 - develop a whole new board, rather than a daughterboard for the pi4 compute module. If I had to guess, it's probably about reducing e-waste: if someone wants a compute module and it's out of stock but there are pi500s available, why wouldn't they buy a pi500, strip the compute module out and dump the rest of the machine in the bin?
And to be fair to them, they're dogfooding a fair amount - the RP1 and the RP2 are in the design, rather than a generic southbridge and keyboard controller.
