Unity's analytics can tell them how many installs there are. It will only trigger an install if someone actually opens the game. So downloads, without opening it, does not count. I have a game published through Unity's UDP that have many downloads (shown by the store) - bots I assume - but Unity shows there's only 1 person who actually opened my game (I assume the tester). I had queried it with them, and they explained how their analytics works.
In contrast, Unity has no way of knowing how much the game actually earns. They will have to extrapolate that from x installs with this funding method is likely to yield x return.
Note the analytics and UDP, and their own marketing division, gives them a lot of data, and they might have implemented their pricing plan based on the data they have. This data most likely excludes small developers, but the price point of $200,000 also means most of us will never need to pay for the software.
I can foresee a scenario where someone earns on average 10c per install, and now faces a bill of 20c per install. But as far as I can determine that's only after earning $200,000, and not retrospective. Their pricing shows it is a cost per new install. So the developer at that point would need to change their pricing model or talk to Unity about it.
What concerns me is everyone is looking to currently "free" software as an alternative. That software is not free, someone is paying for it in the background. It will come down to your risk tolerance - I see using "free" software as too risky - if something happens to the sole developer, can the software continue? I don't foresee most of my games ever reaching $200,000 in earnings and I'm excited about some of the new tools Unity is developing. So I have no reason to abandon Unity. But this is a personal decision every developer must make for themselves.