Not unquestionably idiotic as the vast majority of projects use it, and most people don't even bother looking at it.

Combining your two strains of thought consider this. A small project with only a thousand users has 999 users who only return one result every day, but one user returns 1000 results at the same frequency. How does seeing that all one thousand users are active help when half the results returned in any multi-day time period are actually generated by one user, whereas using the users' RACs will show this very clearly.