The allegations in Ohio that the machines that tabulate the votes have a bias can be easily checked, and this is going on with the Ohio recount.
Actually, no. In many counties, the elections director is cherry-picking the sample precincts to hand-count to check the machines, rather than choosing them randomly, as required by law.
They are supposed to recount by hand if the hand- and machine-counts don't match in these precincts. In practice, counties are replacing the machines and trying again until they get a match, then doing the entire recount by machine.
If the machine is crooked, it is going to have a biased result if the precinct is cherry-picked or not.
And even if cherry picking was evidence of fraud (which I do not accept), this means that it would take many people to be in on it. The OP was wondering how many or few people it took to steal an election. If it takes elections directors in many counties to cherry pick precincts in order to subvert a recount, this is evidence for the "many people" position.