How can you be sure the estimated max server capability is not actually just a limitation in the client, i.e, the client maxes out at sending 224k requests / second.

I see that this is clearly not the case here, but in general how can one be sure?