Why doesn't someone just try to measure this next time!?

Can't really measure without being sure you aren't being messed around with, when it's a remote platform. Stupidly easy to detect when people run such benchmarks/tests against you as well.