As today's Morning Election Digest on the VA-AG race points out, we may have some sort of manual recount of optically scanned ballots in that race. I'm not familiar enough with the new law to know whether a manual review of every ballot in every county is something either side can demand in a recount, or whether that side would have to show some cause to do a full hand count. But whatever the rules in this specific case, in general we shouldn't let what happened in FL 2000 create an unthinking enthusiasm for hand counting in every case.
This AG race is an excellent example of the problem of hand counting a hyper-close race. If you look at the VA state website cited in the item for the AG results, then go to local results, it will tell you that Fairfax County reported 303,083 votes in the AG race. So far so good. But go to the Fairfax County website, and they give you this crucial extra bit of information, that total votes in Fairfax County were 311,429. To be sure, many of those 8,000 "missing" votes are due to drop-off, the fact that some people only go to the polls to vote for the top line race/s, and don't bother voting further down the line. But the state site will tell you that only 306,161 voters from Fairfax voted in the governor's race. So that's over 5,000 votes unaccounted for.
To be sure, some of the 5,000 are folks who came in and voted for some race further down the line, but skipped the governor's race. Maybe they hate McAuliffe. But most of the 5,000 are almost certainly people who thought they voted in the governor's race, but made a mark the optical scanner didn't register. The scanner will kick your ballot back if you vote more than one candidate in any race, or if you don't vote for anyone in any race, but it accepts your ballot if it can't read any vote in a given race, because the assumption is that some voters don't want to vote all the races.
Now, the AG race has over 8,000 "missing" voters. At least 3,000 of those are almost certainly legitimate drop-off, voters who came to vote in the governor's race, but didn't care to vote for either AG candidate. But that still leaves approx. 5,000 ballots in Fairfax County on which the voter made some sort of mark, but that mark didn't meet the optical scanners' criteria for a vote that could be counted.
Bottom line, if there's a hand recount, in Fairfax County alone, there are probably close to 5,000 ballots that humans are going to have to interpret because the scanner couldn't call them. We can only hope that the optical scanners were 100% accurate when they did make a call, or that's who knows how many more votes up in the air.
Now, you may think that's all to the good, because you've got FL 2000 on the brain and imagine that more people who failed to mark their ballots to optical scanner standards were probably our voters, therefore our side will do even better if at least some of those votes can be legitimately counted after human inspection. But a lot has changed in 13 years, and the VA is not FL. People more likely to have Alzheimer's are now more likely to have intended to vote for Obenshain, and worse, the humans who are going to be examining their ballots out in the deep red boonies are not going to have nearly as much scrutiny as folks in Fairfax County looking over the ballots.
The paper audit trail is very important as a safeguard against intentional or accidental systematic error. Computerized counting and tabulating is necessary to get a precise result, but such systems are inherently "black boxes", you can't see what's going on inside them, and they are therefore vulnerable to large, systematic errors. And, of course, the outcomes of elections are so important, that we have to worry a lot about systematic errors that aren't really errors, but crimes.
But the paper audit trail is like Kryptonite to a close election. Humans can't tally 2 million of any sort of object to the point that a variance of 164 means anything. Then throw in the subjectivity of human interpretation of the marks on the ballots, and you can simply forget about the resulting count as any sort of indicator of which candidate actually had more votes cast for them.
FL 2000 was both hyper-close, and it had all sorts of probable systematic sources of error -- the butterfly ballots and "Jews for Hitler", defective and overloaded punch machines, etc., etc. A hand recount of counties where there were these systematic problems was reasonable, and would have yielded a more accurate count. Such a recount might have put Gore far enough ahead that the margin would have been convincing as something beyond a variance caused by imprecision of counting, so it's reasonable to think that should have happened in FL 2000.
VA-AG 2013 is not FL 2000. There is no theory of a potential systematic error that either side is claiming. There were only five races on the ballot, and only the governor, and, in some cases, the local Delegate race, had more than two candidates. The result was a very clear ballot, with all races clearly demarcated off in their own space, so that we didn't have either butterflies or any reason to have Jews voting for Hitler by mistake. There were no ballot punch machines that got clogged preferentially in blue precincts to skew the results systematically in one direction.
Counting the ballots by hand, in this race, is just going to open the gates of Random Drift Hell.