My careful test showed that the Alpha 6000 loses 0.2084 seconds of sound when it changes files, which is good to know in general but not relevant here. That is, the last bit in the file is silent, but the next file starts with the correct synchronization for the next frame. If you concatenate the streams, 1/5 second of sound is missing but no other glitch is apparent. (I did not look too closely at the video frame timing.)
It also shows that over a 30 minute recording it does drift slightly, less than a frame. Measuring, I see the tick on the Sony's audio track appears to the right of the same position on the Zoom H4N track: 0.0055 seconds after about 10 minutes (end of one file), and up to 0.14 seconds after 27m18s, near the longest a single "take" can record on the Sony.
In comparison, the Canon 70D audio track shifts to the left by 0.0065 seconds over the same half-hour span: about 1/3 the amount of drift, and in the opposite direction.
The real problem with my synchronization was not due to the Sony camera. For some unknown reason, the Zoom H4N seamlessly dropped close to 3 seconds (2s19f out of 24 fps) in the middle of the file.
That is, the recorded WAV file (one file—not a file break) somehow stopped laying down samples and resumed a moment later with no indications that anything took place. Looking close-up in Audition there is a line visible in the frequency display where the discontinuity in the waveform is interpreted as noise on all frequencies. One person is talking before the break, and another person is talking after, as if a section was deleted from the recording and saved back to a WAV file.
Strange, isn't it?
