I was looking forward to the hand recount of the razor-close Senate race in Washington's 17th Legislative District, not because I expected Democrat Tim Probst to jump into the lead (he didn't) but because hand recounts are an opportunity to measure the accuracy and integrity of our ballot tabulation systems.

Out of 58,994 ballots cast, Republican incumbent Don Benton now leads Probst by a mere 74 votes, down a net four from the original tally. Probst picked up eight votes from the hand recount, Benton picked up four. According to Clark County Elections supervisor Tim Likness, all 12 of these new votes came from ballots "where the voter made very light marks and/or used a pencil to mark their ballot." These were ballots on which the optical scanners did not detect a vote.

"This is one of the reason for there being a manual ballot count when an election contest is very close," Likness explained via email. "The human eye did indeed identify these as valid votes even though our voting system could not."

To be clear, that's an adjustment of only 12 votes out 58,994 ballots cast: That's a 0.02 percent tabulation error rate. Hand recounts in legislative races are mandatory when the margin is less than 150 votes and 0.25 percent.

Nine additional ballots have been sent to the canvasing board that Likness says either contain marks that are "so light as to possibly not to be considered a vote," or on which voters changed their vote in a way where voter intent could not easily be determined. But it wouldn't be fair to categorize these as tabulation errors.

Every election season I hear conspiracy theories about elections fraud and crooked tabulation software. But on top of all the routine audits, these occasional hand recounts are the ultimate test of the accuracy and integrity of our ballot tabulation systems. And it's a test that no Washington election has ever failed.