That’s assuming that the total voter distribution is 50% for Trump and 50% for Harris. Even if you make the assumption that 99.99% of the total votes went to Harris, the probability of randomly selecting 138000 votes for Harris is about 0.0001%, assuming you put the vote back into the pool after counting it. In reality, you are reducing that 99.99% harris pool each time you count a vote for Harris. For example with a total pool of 276000 votes, that 99.99% degrades to a mere 50% chance of pulling a Harris vote on the last draw. In other words, even if 1 in 10000 people voted for Trump, the chances of pulling all Harris votes from the counting pool is literally, and drastically (multiple orders of magnitude) less than 1 in a million