[tmaly]

If you look at it, the change that was most attributed to the slow down, was committed on October 2016.

Why could the people making an issue about the 0.5 us slow down per request not have tested or ran a benchmark sooner?