You mess around with the Twitter API just for kicks to build a bot tweeting every time someone orders a pineapple extra-cheese pizza as an in-joke with your friends, and then, hey, you know how to use Twitter's API, and building that expertise helps Twitter.
This is the same reason why Linux owns enterprise IT. Not because it was the best solution compared to commercial alternatives, certainly at least not 25 years ago. It's because everybody ran Linux at home, they messed around with it for fun, and they learned it. And when they develop software they'll build it on Linux and thus it's easiest to deploy on Linux too.
Breaking the free API was pretty stupid. Not an incomprehensibly huge self-own like breaking third-party clients (which triggered tons of people I follow to move to Mastodon) but in the same zipcode.
And people were _largely_ assuming this for non-critical but fun/useful things, like the aforementioned bots, or Threadreader, and so on. So the stakes weren't _that_ high.
We're seeing that assumption fail now, of course, but arguably in a pretty weird difficult-to-predict way.
I think the businesses expecting Twitter to remain relatively stable for the next several years were the reasonable ones here, yes.
I think that without a free API the service will become far less relevant. Sure, you could monetize access to it, get a few cents here and there. But far fewer tools and sites will integrate with Twitter from now on.
Twitter needs money, yes, but I doubt they will grow again with decisions like this.
It seems like Twitter is already backtracking on this, but the damage is done. No one wants to build for a platform that doesn't understand the value of builders.
That's from the outside looking in, though. I never made a Twitter account, I don't know if their EULA promised free beer and cookies ad-infinitum.