This is it, and it's the same with Twitter. At some number of connection in the social graph, or some amount of content produced, a user becomes more valuable than the ad money they could bring in. i.e. the opportunity cost flips, and it's worth giving up the ad revenue or API usage in order to keep them. To use an extreme, if a Kardashian said they were leaving Twitter, it would obviously be worth a lot of money to keep them on the platform. But my guess is that the percentage of users bringing more value than their ad revenue is closer to 1 in a 100 than 1 in a million.

As you said this then plays into the third party client issue directly, because those users are almost by definition power users, and power users get so much value out of third party clients with micro optimisations for their use-cases.

The problem is that this feels so obvious that I can't believe Reddit (or Twitter) don't have a measure of this internally, and I don't know why they wouldn't be optimising for it. My only conclusion is that it's too much nuance for a Musk-driven product team to handle, and that Reddit are shit-scared that they're going to collapse before IPO'ing and can't make rational decisions.