The webservice memory limit and request recommendations were based on
Omnibus defaults for Unicorn. Now that we default to Puma and done load
testing with it, we can update these values to match what was observed
during testing.