HTTP.jl doesn't seem to be good at handling over 1k concurrent requests, in comparison to an alternative in Python?

the first http.jl test has been added to the Techempower test;
( and the final PR deadline 12/28 ; if anyone wants some tuning )