Because the incentive is now there. Maybe they don't get enough paid customers and want more money. This puts a bit of pressure to move a new feature that is really handy into the paid level. Then another and another. Might not happen but it could.
Most people using Datastar will not necessarily be smart enough to fork it and add their own changes. And when Datastar makes a new release of the base/free code people will want to keep up to date. That means individuals have to figure out how to integrate their already done changes into the new code and keep that going. It's not a matter of if something breaks your custom code but when.
Finally, many people internalize time as money with projects like this. They're spending many hours learning to use the framework. They don't want to have the effort made useless when something (ex: costs or features) changes outside of their control. Their time learning to use the code is what they "paid" for the software. Doesn't matter if it's rational to you if it is to them.
arguably... then I'll argue the opposite (even as a subscriber). Jetbrains rolls your subscription back to the version on the day you purchased, not the last version in your subscription. If you purchase and immediately find a bug, you know you're rolling back to that bug unless you pay a renewal next year. Same goes for any fixes you got during that entire year. Poof, gone.
I had always thought the same thing about high use by aging out coders. Don't know how representative it is but the just released Go annual survey results seem to show that isn't the case.