Statistically anomalous data exists, it should be scrutinized, just partially. Sample and do a hand recount for 1 district that shows the greatest evidence of bullet ballots, if that turns out close to expectations, that’s the end of it. If it isn’t… that’s a massive can of worms.
1 anomaly can be seen as a random outlier, and if it wouldn’t impact the overall result, can be ignored. Multiple anomalies, in only the places that matter is worth a little time to confirm.
Jury nullification.