I don’t do caffeine unless I’m desperate, but Friday was a mind-numbingly boring day, and Diet Coke was called for as I observed a full day of the recount in the Attorney General’s race and the Fifth Senatorial, where Robin Wright-Jones had edged past Rodney Hubbard by only 101 votes.

First, I observed workers (always one Democrat and one Republican) take turns going through paper absentee ballots. One young woman used a rubber thimble on her middle finger to help her pick up the pieces of paper one by one. (Everybody else dipped two fingers in a pink paste called SORTKWIK to make their fingers sticky.) While her partner watched, she either put a ballot in the pile of those correctly marked or, occasionally, in the pile reserved for screwups who can’t fil in a bubble with a pen. That latter pile would be hand counted later.

Apparently I missed what minimal “excitement” was to be had when, on the previous two days, the hand count of paper ballots cast on primary day took place. Some discussion of voter intent occurred in the case of a few ballots. Perhaps four or five of them are still in dispute–a far cry from anything that would change the result in either race.

On the day I was there, the ballots that had been filled in correctly were taken to another room and fed through an optical scanner (by one Republican and one Democrat, of course). The ballots were then packed up in boxes of five hundred, sealed, and initialed by both people.

At other tables, pairs of workers were tallying the votes on what looked like long rolls of cash register tape from touchscreen machines. They sat across from each other with a long four inch wide board that had a couple of metal brackets screwed into it eight inches apart. They pulled the tape under those brackets. One set of votes exactly fit between the brackets, so one worker would pull the tape through to each new set of votes. His partner would read the votes in the two races being recounted, and the one who pulled the tape toward himself would record the votes on paper.

At the end of each tape, the workers compared the number of votes listed at the end of the machine’s paper trail with their own handwritten tally. What they hoped for, and usually got, was a match. On one occasion where the paper tally was off by one vote, they had to run all that tape under the brackets and recount. Fifteen minutes. Bummer. The young man whose count was off flushed at his mistake.

Of the half dozen observers in the room, only one was actually attempting to write down the totals each pair of counters came up with. That was a man who was there on behalf of Wright-Jones. He pointed out to me that 270 rolls of votes would be counted in the Fifth Senatorial race. Since only 101 votes separated Wright Jones from Hubbard, a one vote difference on less than half the rolls could change the outcome, so he was keeping a close eye out.

You know I don’t trust touch screen machines or op scans either, for that matter. But the people running the recount and those doing the counting acted as if their reputations were on the line. They answered any questions we had and were scrupulous about following the rules. Although one pair of counters miscounted the same roll three times in a row, everybody else remained admirably alert despite the monotonous task at hand.

So human error was not much of an issue. Machine malfunctions cost the workers far more of their time. An indecipherable ink mess caused by a paper jam could occur near the end of a roll, necessitating that the roll be reprinted and recounted. For example, at the end of one particularly huge roll, an ink mess threw the count off. The hand count showed Koster one vote short of the machine tally. Since a vote was obviously hidden in the ink mess, and although that vote was “almost surely” for Koster, they reprinted the roll to be recounted. Sigh.

One pair of counters ran into four straight rolls with printer jams. SIGH. I would put the number of rolls with ink messes that necessitated a reprint somewhere around 15-20 percent.

Another complication was that occasionally a printer had stopped and later started again without making an ink mess. While the printer was stopped, people voted and the machine counted those votes even though the printer failed to record them. The only way to know when that had occurred was when the hand count produced fewer votes than the machine said it had counted.

The recount was obviously being run on the up and up, but I mentioned my dislike of voting machines to Kyle Dubbert, who seemed to be in charge. His official title is “Policy Supervisor” and he’s a Republican. He said that if enough audits are conducted, then he would catch any widespread discrepancy between what the machines tallied and what the printers printed out. He  told me that the city conducted more audits than the state requires. The state requires an audit of one precinct for every 100 precincts. The city conducted five audits out of 203 precincts, which is about 2 1/2 percent of the precincts. Kyle said that if they’d had more workers, they’d have done more audits in the two weeks they had for that.

The other possible kind of fraud that concerns e-voting activists is that the machine would record a different vote than what was intended as well as print a different vote than intended. Kyle felt that the only protection against that possibility is for voters to examine their printed receipts, and he said the city urges them to do that. Of course, considering that many people are careless enough to vote in downticket races on whether they like the sound of a name, it doesn’t seem likely that most of them will double check voting receipts.

In any case, we’re stuck with the machines, and the recount is proceeding even as I type, probably. I see no indication that either race is going to come out any differently.