With the final votes being counted in California’s 21st Congressional District and Democrat T.J. Cox poised to oust Rep. David Valadao (R), we appear to be at long last ready to put the 2018 House midterm elections behind us.

But not so fast.

North Carolina’s Board of Elections has certified the results in 12 of the state’s 13 House districts, save the 9th District. There, as we reported Thursday, the certification has been held in abeyance under the provisions of a state statute that allows the board to “take any other action necessary to assure that an election is determined without taint of fraud or corruption and without irregularities that may have changed the result of an election.”

In other words, the board is worried about voter fraud. As it stands, with all votes counted, Republican Mark Harris holds a 905-vote lead, a narrow 0.32-percentage-point margin. The question at hand is whether all of those votes are legitimate.

We often note that in-person voter fraud in the United States is so rare as to be nearly nonexistent. When President Trump asserts that thousands of people changed hats in the parking lot so they could fraudulently vote twice, we dismiss it summarily. So why is it that, this time, the allegations are treated seriously?

1. The numbers look strange.

First of all, there are some weird aspects of the votes that are in.

Eight counties are entirely or partially in the 9th District. Seven of those voted for Democrat Dan McCready over Harris and Libertarian Jeff Scott (who earned 1.8 percent of the vote). The two outliers are Bladen and Union counties.

In 2016, Trump won six of the counties. In every county, McCready performed better than Hillary Clinton two years prior.

Bladen County was the only one where the number of votes for the Republican in 2018 relative to the number of votes for Trump was well above average across the eight counties, while the number of votes for the Democrat relative to Hillary Clinton was below average. In Union County, the other county Harris won, votes for Harris and McCready were both above the average across the eight counties.

There’s another way in which Bladen County is odd. In every other county, mail-in absentee ballots favored McCready, the Democrat, by at least 16 points. In Bladen, Harris won by 24 points. He also won early voting (that is, in-person voting), but that happened in Union County, too — where Harris won the county but lost the mail-in ballot by 20 points.

As a percentage of all absentee votes (early voting or mail-in) cast, Bladen’s mail-in total was well above any other county’s. About 1 in 8 of those votes were mailed in.

Others have noted different oddities. For example, Michael Bitzer, a professor of politics and history at Catawba College in Salisbury, N.C., noticed that nearly as many mail-in absentee ballots came from unaffiliated voters as from Democrats. Across the district, Democrats were 10 percentage points more likely to cast ballots than unaffiliated voters.

From our report:

“Gerry Cohen, an election law expert who used to work for the state legislature, said he found one precinct in Bladen County in which the results seemed odd. In that precinct, called Bladenboro 2, 159 people voted by mail — 18 Democrats, 32 Republicans and 109 unaffiliated."

2. There are reports of unusual activity.

There can often be seeming statistical oddities that crop up in elections. A number of people point to precincts in Philadelphia where Mitt Romney received no presidential votes in 2012 as evidence of fraud. (Trump himself referred obliquely to this during the 2016 campaign.)

But there wasn’t any accompanying evidence of people having actually committed fraudulent acts. Ryan Godfrey, who worked as an election integrity official for the city, noted that the Philadelphia Inquirer went to find Romney voters in those areas — without luck.

In Bladen, there’s evidence of weird activity.

“At least five affidavits submitted to the state board described various instances of fraud, including multiple occasions when people came to voters’ doors to collect ballots and offered to fill them out for them,” our Amy Gardner and Kirk Ross reported.

The Charlotte Observer detailed several specific examples.

3. This is the sort of alleged fraud that experts actually worry about.

Perhaps the most important differentiator here, though, is that mail-in voter fraud — unlike in-person fraud — is actually something that experts worry about.

Most voter fraud laws target voter impersonation at the polls -- the idea that people show up to cast a ballot in person and must therefore confirm their identities. In 2014, I wrote about the disconnect between laws targeting alleged fraud and the actual problem. I exchanged emails with Justin Levitt, a professor at Loyola University Law School who has done research on voter fraud for the Brennan Center for Justice and who wrote an article documenting only 31 examples of in-person fraud in millions of votes cast.

“Levitt made clear that absentee ballots can be a threat to the integrity of elections,” I wrote at the time. “He pointed to instances in a Pennsylvania state Senate race in 1994 and the Miami mayor’s race in 1998 as examples. Fraud in absentee balloting is ‘unfortunately quite real,’ he said.”

That’s what’s alleged here.

So we have numerical evidence of possible abnormalities, anecdotal evidence of abnormalities and alleged abnormalities targeting a type of voting known to be subject to possible fraud.

That’s why this case — even if the irregularities didn’t change the outcome of the election — is worth worrying about.