While discussing the rollout plan for query stats, we considered different levels of rate limiting. We don't have great data on this. To help make decisions like this in the future, it would be great to run our high value benchmarks on a couple different rate limit variants, so we can summarize the expected impact better. As a start/guess, maybe we should have: 1, 10, 100, 1000, 10000? 

100 already exists.