Now that I've been forced to recall all of this to a greater detail, I think can actually tell which of the drives produced the bad file: it's the Hitachi drive (HL-DT-ST).
If it turns out that it's as if the left sample of the first stereo pair was omitted and the stream began with what was supposed to be right sample then I'm wrong and the problem is with the other drive, so a more monaural portion of audio will still be very helpful. I can explain how a CRC that doesn't include null samples will still be the same in this situation, but it's largely academic.
A heart-felt thanks to Surfy!
