This A.P. update was posted not long ago on the Newsday site, which has been following this story about the election for state senator in Yonkers:
"With most machines recounted, Spano trails Democrat"
http://tinyurl.com/59kwo
to summarize: Democrat Andrea Stewart-Cousins was down by 1,674 votes before a "double-check" (yes, that's the word the story used) of the voting machines in her district. With the "recanvass" mostly finished, she's now ahead of Republican incumbent Nicholas Spano by 263 votes.
The original numbers: Spano winning, 51-49, with 54,120 votes to Stewart-Cousins' 52,446 votes.
They still have to count thousands of absentees and provisionals, too, so the race is still in question.
but what isn't in question is that the original numbers reported by the elections office have turned out to be very wrong.
It's unclear to me exactly what they were "double-checking" -- it doesn't sound like paper ballots. Here's what they said in yesterday's Newsday story:
"As the recount progressed on Monday, election workers, closely watched by representatives of the campaigns, compared vote totals recorded on machines with the tallies reported by poll workers on Election Night.
"Jonathan Rosen, director of the state Democrats' campaign committee, said Berger told him that nearly every machine had a higher count for Stewart-Cousins than was recorded. However, Spano also gained votes, including 80 from one machine, Rosen said."
and another hint of mystery:
"The county Board of Elections disclosed Friday that seals on 22 machines were damaged or missing."
plus:
"The two sides have traded allegations about irregularities in the voting, with Republicans claiming voters were imported from New York City and Democrats claiming voters were harassed and voting machines were tampered with. All machines and records were impounded by judges.
http://tinyurl.com/6fyhj