I'm lost why your doing test using WEBrick. Unless you plan to use it
as your production server you're likely to get a completely different
profile of results using Apache.
You've proven that there is a difference between performance of 2.2.2
and 2.3.3 under WEBrick. The question is what does that mean? That may
be reflected in Apache or Nginx or the results could be completely the
opposite. WEBrick results could be a hint as to what happens in real
production servers but you don't know until you try you benchmarks on
real production servers.
Also, you benchmarks need to reflect your actual workload but as Opera
would say 'thats a whole nother show'.