I always interpreted this as someone wrote a script and learned to game the algorithm, you think there’s a line of code in there that says trump_replies.unshift(crazy_doctor_brother_tweets) in there somewhere?
Popular non-'progressive' senders - Trump and now Musk being the poster child of such - probably get their own comment manager who makes sure there is always some vitriol waiting no matter the message. This does not scale but the Pareto principle makes that it does not have to since there are not that many of such. The same may be true for popular 'progressive' senders where comment managers may make sure there are supportive comments but this seems to be less clear-cut.