I think that due to how sophisticated anti-bot measures have gotten, bots now go through a "life cycle." An engagement bot spends the first phase of it's life building up an innocent and legitimate looking history. It does this by blending into the noise with innocuous and pointless comments that you'd never take a second glance at, and definitely not report or flag. Then when the account has aged and is in good enough standing, metamorphosis to the adult stage occurs, and the bot starts posting the kind of blatant spam that you'd think would be automatically ban filtered, but somehow isn't. These bot farmers are quite literally farming bots like vegetables and selling them when they've ripened.

I am very confident that this is the case in YouTube comments, where most people find they can not use violent words like "kill" or "genocide" when discussing war, but somehow there are bots posting uncensored racial slurs.