Looking at these graphs:
https://chromeperf.appspot.com/report?sid=0a000d48f4a6fcafd84aa23b4d8ad180e4a9d6b6f61775305b14793f0569a5fb&start_rev=386972&end_rev=416946

and these graphs:
https://chromeperf.appspot.com/report?sid=18263772de0fb116e92aa17c83e5e66b374e7b482a6e2c3bee7558e246437e45&start_rev=390869&end_rev=393562

It looks like we had a regression. I kicked off some perf bisects here:
https://chromeperf.appspot.com/buildbucket_job_status/9000964595479350416
https://chromeperf.appspot.com/buildbucket_job_status/9000960785830955632

Which gave me this change to the make time code:
https://chromium.googlesource.com/chromium/src/+/89c5c64c72a737a52426df0f29146fb530b9129c
(diff https://chromium.googlesource.com/chromium/src/+/89c5c64c72a737a52426df0f29146fb530b9129c%5E%21/#F0)

I'm guessing that either we're measuring the new addition perf cost (or maybe a de-optimization of that method), or the time is now more accurate. Adding respective parties so everyone is in the loop, but I'm guessing we'll mark this as WAI.

Eric, I was told to loop you in as you might have info here.