[01:02:25] <Krinkle>	 ori: im a bit distracted at the moment, hadn't noticed! One candidate that comes to mind is also valetin's work on upgrading ATS and fixing an origin coalescing bug that was causing significant stalls at the traffic layer
[01:02:55] <Krinkle>	 In extreme cases a cut of 100ms for one request
[01:04:05] <Krinkle>	 I imagine for a p75 metric of several roundtrips (html + css), that could very well eg have experienced something like 10-20ms impact that has a 100ms impact there
[01:05:31] <Krinkle>	 Looks like median responseStart is down by 45ms or some such