One thing I don't really understand... in the recent interview Dave said that audio input on a poly synth doesn't really make sense because of where to send the audio when you have multiple voices. Yet there's audio mod. Does that mean the audio modulation is on a per-voice basis?
Yes... it modulates on a per voice basis.
Personaly I don't see the greatest of uses for this source... I think DSI simply put it in there because they can. I bet that the signal is taken at the last possible digital spot in the signal chain which would be right before the digital delays, and that this signal is probably mixed digitaly with the output of the delays, before being sent to the Audio Out mod.source.
It's a little funny though, that Dave talks about polysynths and Audio Input as nonsense... he did this himself with the Poly Evolver... it routes the Audio Input into all four voices.... the new KORG Minilogue also does this. But I give him, that I don't really see the use for it

... especialy if there is not a separate input for each voice... the later would allow for some use if in a multitimbral invironment, as each voice would essentialy be a stand alone monosynth with Audio Input.