Having dabbled in the space a bit: (on the less-malicious side, scraping for archiving, I imagine pro bot-writers are far beyond me.)

There's an upper bound to "bot complexity"; that being the complexity you want to impose on a normal user. For many sites (e.g. instagram) it's often easier to just "pretend to be a normal user" and hit it from a headless browser. You lose some functionality, but at least for my needs, that's always been a sufficient fallback.

As such, I've always been of the opinion that the primary function of API limitations is to punish/limit/extract more money from the good participants.

Sure, it affects things at the margins. However that is all this does. The biggest outcry against bots is not the low margin spam/advertising related stuff. People learn to ignore that type of content almost as soon as they start using the internet. The real troubling bots are those used for harassment, influencing financial markets, and influencing political discourse. The first was never about pricing. The latter two are still amazingly cost effective. I would bet they would have to go up by an order of magnitude or two before the pricing effects start having a real impact.