So drift between cameras is not an issue so long as the frame rates are set correctly.
What of clock drift from the double-system audio? Since that's referencing samples per second and not frames per second, I imagine there would be some loss of sync over the course of an hour.
Is there a way to derive sync from an SDI signal? Methinks if I could get genlock from this and turn it around to word clock, I'd be able to have my audio be sample-accurate with the video.