Well... I think your conclusion is absolutely right, but it's actually more complicated than that.

Executive summary of the math from the conversation to which I alluded: If all things went absolutely perfectly for Selig, he couldn't possibly have made more than about $450K per year, (gross) at the peak of Apollo's popularity. The app was around for less than six years. Therefore, excluding expenses, his maximum theoretical gross income was still somewhere short of about $2.5 million. (There was a lot of math stuff that led to that figure.)

Reddit is demanding fees essentially equivalent to seven times that figure... per year.

That math doesn't add up -- unless you assume that those existing third-party devs aren't the target audience at all, for that proposed fee structure. At no point did Reddit ever think that these small time devs were going to be able to cough up those exorbitant fees; the target audience is and always has been large language models.

The thing is, I don't actually think there is much chance that the LLMs are going to be any more likely to accept the new fee structure than those third party add on developers. Thus, as you've stated, there will ultimately be no payday for Reddit.