> assuming even distribution

I don't work for Uber, but this is almost certainly the assumption that is wrong. I doubt there is just a single workload duplicated 2.1K times. Additionally, different regions likely have different load.