It probably wasn't the hardware so much as the lack of time. The Mac boot-beep already does a hacky approximation of lowpass filtering, and I can think of a couple other quick-and-dirty ways to mimic analog-style subtractive synthesis that would be passably performant on an 8MHz 68000, but according to Hertzfeld he only had about two days to do the entire thing, so I imagine at that point they just had to make the call not to pursue it further.