Suricata

Have been chasing this for a bit.
I have a reproducible case repay with trex (configs are available in the internal OISF QA lab) where simply removing commit https://github.com/OISF/suricata/commit/5c95c28387a4a756e39fd5f2c984bb8b034aa485 results in what it seems better performance (10-12% less drops)
This may or may not be expected/intended , so it needs more investigation. However just removing the commit and redoing the same test immediately result in better perf stats.
It is especially visible if a box is optimized for max performance with the current HW and then the commit added (suri recompiled and the test re-run).