I'm waiting for the comparison between exit polls in precincts that used touch screens and those precincts that did not. If we can show that historically exit polls are accurate predictors, then we can show a sliver of evidence of blackbox errors.
Of course, the CNN exit poll is useless because of the late night "correction".
I see some of the data here : http://www.dailykos.com/story/2004/11/4/212931/005.
In Nevada (anecdotal evidence at this point) they used a 100% computer system with paper verification and the exit polls were accurate within .1%. In Ohio, they used punch cards and the spoilage was high in Kerry precincts and the exit polls showed Kerry winning.
I'm going to keep digging but if someone has data, please beat me to the punchline.