The answers to your leading questions are probably classified at the TS/SCI level.
I think it's safe to assume that there are a non-trivial amount of psyops happening on those platforms, since those sites are among the main sources of information for almost all people. It would be foolish to not target such low hanging fruit to manufacture consent if you are trying to influence populations.
I would go further and add wikipedia and google results to the list. Conflict of interest edits on wikipedia by companies and legislatures are nothing new. Check out CongressEdits[0] and the slew of bots that it inspired, which enabled monitoring of IP address ranges that edit wikipedia.