Github user joewitt commented on the issue:
https://github.com/apache/nifi/pull/2416
also of interest am now running both before PR and after PR codebases on
otherwise identical configurations and the new one continues to have a very
slight but consistent performance advantage. Of note is that there is less GC
time on the new one compared to the old one by about 50 young gen sweeps and 7
seconds of total time.---
