Yes, it probably is.  What is jitter recording?

What's also odd is that this was previously landed in r401796 and reverted in r402294 (June 23rd to June 27th), but the graphs don't appear to have changed in that spot.

Re #1 : Jitter page set has layers that are scrolled on compositor and there is javascript that undoes the scroll. So on every commit, the scrolling is undone on those layers and they move back to their original position. The metric runs on those pages and measures how much we have scrolled on the active tree before it was undone.

Hmm.  On OSX (but not on retina) I tried reproducing this, but was not able to.  The value of the jitter metric locally for me appears to be roughly the same at r408172 (my patch), r407000 (before the jump), and r409000 (a day or so after).  That said these results are pretty noisy even when I run it a few times, so it's hard to tell.

Do you have any suggestions for reproducing this locally?

Deleted: Screen Shot 2016-09-13 at 3.38.46 PM.png 129 KB

	Screen Shot 2016-09-13 at 3.38.46 PM.png 129 KB View Download

Hmm, I can't think any easy ways to reduce the noise(except just increasing the repeat count). If you are getting the same values before and after the jump, maybe this is retina specific ?

I was unable to ever reproduce this and unified begin frame is not going to unship at this point.  Closing, sorry.