Remi Broemeling

@mariussturm I thought about that, but I don't see how that could be the problem if we assume that a record is never "half-accounted" before the metrics are sent. By...

Hmm. Ok, thanks -- this is rather disquieting, honestly. I think that what we'll do is remove this plugin from use and change to doing elasticsearch queries ("after the fact")...

@mariussturm I never tried setting outputbuffer_processors to 1 to see if it would fix the issue, I was under the impression that you didn't recommend that as a "real" fix....

@mariussturm I turned outputbuffer_processors = 1 (from the default of 3) at 12:13 mountain, and from 12:38 mountain through to 13:36 mountain, I have 5 instances where upstream_response_time is greater...