Choosing installs to collect revenue on is stupid because it isn't tightly correlated to revenue for a game:
Demos, bundles, free week-end tryouts on Steam, free-to-play games and other specific cases all lead to high install counts without necessarily generating the matching sales.
It's especially stupid because Unity has good market share on mobile, which is a juicy market but with a fixed price per install, free-to-play and low-priced (0.99$, etc...) games will never choose Unity for new projects, because the percentage of revenue you will lose on those is going to be prohibitive.
A less worse metric to collect on would be actual game sales count, but that's still horrible for promotional sale events and low-priced games (assuming fixed cost per sale).
The sensible and simpler option would simply be a percentage of revenue past a threshold like Unreal Engine and CryEngine for example.
I see two possible outcomes:
* They stick with this pricing policy, and cash in on the fact that a lot of game studios are heavily invested in Unity at the moment and will take time to move away to extract a shit-ton of money short term in the meantime. Long-term it will probably crash and burn, but they're fine with it as long as they can milk enough before that.
* They wait for the outcry, say they've listened to the community, roll it back and introduce a more sensible royalty system with percentage of revenue, which the community grudgingly accepts since it's less worse than the initial option.