Thanks for the link. I've now read the whole thing and I don't see any holes in the arguments. The author does not make any definite conclusions, but I'm at least now reasonably satisfied that he was not using early exit poll data. (Assuming that the source had been updated regularly, that is.)