I think the reasons for the discrepancy are obvious: Much in the way that Reddit will fuzz vote counts, the public like/dislike counts will stay up on accounts

I think the reasons for the discrepancy are obvious: Much in the way that Reddit will fuzz vote counts, the public like/dislike counts will stay up on accounts that are shadow banned and get removed in batches so that the accounts which are bad actors will not know for certain if their vote still counts or not. So what you've documented here is evidence of which videos are most botted.