It's easy to make an enormous query by joining to other data (or to the same data), or reading a lot of data.

A regex query on response_bodies would churn through 2.5TB of data every time it's run.