Are you running many others studies without this happening? It would be very helpful to know that this is isolated to a single experiment and that many others are ok on the same machine. In any case, I'd like to see a copy of the entire experiment if possible--including your MediaLab files.