Can you try to spend 30 seconds countering whatever problems you came up with and only then propose non-trivial problems? I promise I gave the idea more than 30 seconds of thought and did think of the obvious problems, it is just tiring to exhaustively list out and pedantically enumerate a complete proposal instantiated for this specific domain. That is also why I explicitly said it solves the central 90% and would require tuning and mechanisms to reduce abuse since I do, in fact, know the proposal is incomplete, just that the problems I am aware of with this class of solution generally appear to be quite amenable to solutions that should be at least as robust as the status quo.