As someone who worked on a social media platform (large, but not one of the largest), I got to see the evidence that admins and moderators used to make their judgements (which they aren't given much time to do) and frequently the banned user mischaracterizes their actual behavior (both current and previous) and has an incomplete understanding of what the ToS says or the implications of the literal ToS text.

I don't know about Stallman or more academic exercises, but the average person arguing on social media is very likely to go over the threshold of acceptable behavior and may not even know what the threshold for which social media companies are required to report content to police.

The best thing platforms can do to mitigate some of these issues is to give public explanations of their ToS as the moderators are taught to interpret them (basically reduce the information loss from the ToS legalese) and to give moderators more freedom to explain exactly what behavior violated exactly what part of the ToS (reducing some of the confusion from ambiguity).