Similar alert on CrOS, and looking at a graph there may be a bump on Windows as well? 

https://uma.googleplex.com/timeline_v2?sid=35aaf35ff99e3ab87b0a68b98b4c278b for the platform comparison. (7 day smoothed to hide the weekly cycle).

So I took a look at this on just Windows (figured of the platforms seeing this the data would be best -- highest population + best networks) on all the different channels:

https://uma.googleplex.com/timeline_v2?sid=205a3abd6f8f19a6b969d48fb3f4f529

(I left out canary from this plot because I didn't see a change in it, it was noisy, and it was obscuring the stable data.)

The thing that strikes me from that graph is that there's also a rise on beta, and it's at the same time as the one one stable, *not* at the same milestone.  Which suggests to me the problem is external to Chrome.  Opinions?

These look like they're different.

If you shard the Android one by version and look at when each version was dominant (UMA really does not make this easy to look at), it seems M59 was a perf win and M60 was a perf regression.
https://uma.googleplex.com/timeline_v2?sid=718807cbba3ee685611eca72fdf75ba4

Did something of note go into M59 and get reverted in M60?

There seems to be a pattern where latencies for older versions rise up as new versions start to roll out.  Population bias?  First users to pickup new versions seem to have better performance (I can imagine several reasons why).

https://uma.googleplex.com/timeline_v2?sid=7bdcd9ad0a750a2a700aa4d108cd8afa

At the moment M61 is enjoying better performance, will it last?

As ckrasic points out, a lot of this is explained by population bias. If you look at the versions split out, and concentrate only when they are dominant they are basically the same % noise.

Here's a cleaner graph than the one David posted:
https://uma.googleplex.com/timeline_v2?sid=7c0dcbd1f97e85c79be98e69a32294cf

Pretty sure this is WontFix.