When doing cut & paste, the video tracks use per-frame
accuracy and the audio tracks use per-sample accuracy.
This wreaks havoc on the A/V sync because the audio
sample can be anywhere between the start and end of the
video frame. This means each edit potentially
introduces 1/frame_rate seconds of A/V de-sync.
If both audio & video tracks are armed for editing then
the audio should align with the video. In other words,
translate the frame offset to its corresponding audio
sample (current_sample = audio_rate * current_frame /
frame_rate). Otherwise i have to go through the XML
file and fix up all the offsets manually.