natural vs "artificial" request rate

wrk

$ wrk -s wrk_report.lua -t 1 -c 1 -d 1m http://target.host/test
Running 1m test @ http://target.host/test
  1 threads and 1 connections
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency     0.94ms    2.83ms  38.04ms   94.75%
    Req/Sec     3.24k   677.21     4.40k    65.67%
  193542 requests in 1.00m, 18.64MB read                <-----
Requests/sec:   3225.03                                 <-----
Transfer/sec:    318.09KB
------------------------------
50%,228                                                 <-----
75%,263
90%,1398
99%,16594
99.9%,27629
99.99%,33696

wrk2

$ wrk2 -s wrk_report.lua -R 3000 -t 1 -c 1 -d 1m http://target.host/test
Running 1m test @ http:/target.host/test
  1 threads and 1 connections
  Thread calibration: mean lat.: 285.547ms, rate sampling interval: 808ms
  Thread Stats   Avg      Stdev     Max   +/- Stdev
    Latency   299.91ms  362.21ms   1.57s    87.86%
    Req/Sec     2.92k   278.87     3.47k    68.85%
  175421 requests in 1.00m, 16.90MB read                  <-----
Requests/sec:   2923.70                                   <-----
Transfer/sec:    288.37KB
------------------------------
50%,214783                                                <-----
75%,317951
90%,799231
99%,1550335
99.9%,1569791
99.99%,1574911