Ah, you might be right! Contention is what can be parallelized, right? So with roughly 100-200 requests/s you end up with 1-0.5 contention if I understood tha

Ah, you might be right! Contention is what can be parallelized, right? So with roughly 100-200 requests/s you end up with 1-0.5 contention if I understood that right. That moves me even further towards agarlands points, though - if I plug that into the equation, I end up with >50k qps. The used numbers create an insanely distorted idea wrt real world performance