In the beta phase I seem to recall there was a problem with some of the karmas locking up on long ethernet transfers. If I recall correctly, the problem would only happen on ethernet and not on USB. The problem is that it only happened on certain 10/100 ethernet switches and not on others. I think it was related to a minor timing difference between the switch's electronics and the ethernet electronics in the Karma itself. It was one of the problems that I spent a lot of time working with the Rio folks trying to solve. We went back and forth a lot and it came down to the wire near the end of the beta phase. I don't recall whether we narrowed the problem down and solved it in time for final PCB production, but the fix (again, if I recall correctly) required replacing a component on the PCB with one of a different value. Or perhaps adding a component bridging something, I don't remember exactly. I remember doing this myself on at least one of the beta Karmas, it was a real PITA, the components on those boards are hella tiny.