Initially all the reports involved headsets, but since then we've had 2 where it came through the handset, I don't think any were using speakerphone at the time, but we stopped blaming the headsets and we now blame the phone itself..
We pursued this with the carrier in case something weird was happening on the WAN, although I wouldn't have bothered personally - too much of a long shot, and I don't see how any kind of fault on the WAN could generate actual audio data and insert it into the packets.
One of my colleagues suggested checking out PoE wasn't causing it. My opinion is that it's the phone itself having a problem replaying the audio, perhaps a problem with silence suppression.
Problem is it's not replicable so it's hard to figure out what's causing it.