It depends! There should probably also be a process by which reviewers themselves get graded. Then paper writers can choose whether to splurge for fewer really amazing reviewers, or a larger quantity of mediocre reviewers. Also, readers will be able to see the quality of the reviewers that looked at a preprint.
It might be possible to have a third-party manage the reviewer ratings. Although I suspect some fields are so small/niche that if someone wanted to associate some random ID with a real person, they could match writing styles etc.
Splurging for "amazing reviewers" could also be gamed to "splurge on those reviewers who are likely to rubber-stamp my submission to get paid" (not unlike some of the questionable open-access journals' current business models).
