The GIFs don't capture the effect well, as the frame-rate is too low; the actual jitter is 2 or 3 times as fast and it's headache inducing almost instantly. 

Interestingly, this only seems to happen at certain window widths (e.g. 1738x1208) and shrinking the window makes the problem go away. Perhaps a problem with a % layout operation rounding fractional pixels?

Looking at this with the Chrome Profiler, it looks like a ton of work is going on at "idle", so this may be a site issue rather than a browser problem.

Deleted: ChromeProfile.7z 4.0 MB

ChromeProfile.7z 4.0 MB Download

Deleted: BusyProfile.png 35.4 KB

	BusyProfile.png 35.4 KB View Download

This is very similar to  issue 677686  but that was seemingly Android only. Are you able to consistently reproduce this?

Re #2: Yes, I can reproduce this pretty much at will, sometimes I have to resize the window around a few times.

There are a ton of timers firing every few ms on the page but I cannot reproduce the issue nor get it to trigger painting/layout with anything near the reported frequency.

I'm still able to reproduce today with the latest (57.0.2979.0) when the window is 1738x1208 and 1695x1302.

If I run
document.addEventListener("scroll", function(){console.log('*scroll');}, false); in the console while in the flicker state, I get an endless stream of scroll notifications in the console.

If I use the Pause button in the developer tools to pause script execution, the flicker stops immediately, which suggests that this most likely a bug in the page itself.

The Ad scripts involved appear to be from DoubleClick, so I wonder whether maybe there's a known issue in their code?

Adjusting the document.body.scrollTop property from the developer tools causes the affect to disappear at some values, and appear at others. Swapping between the values is stable (e.g. if document.body.scrollTop=667 causes the effect, going to document.body.scrollTop=669 will stop it, and reverting to 667 will cause the effect again).

I've now reproduced this on a different machine (window size 1730x1049).

eae@ can you help me understand what feedback is needed?

What DPI and page-zoom are you using elawrence? Does it reproduce if you change/reset the zoom rate and/or DPI?

Both repros were at Windows' default 96dpi, at 100% zoom. The instant you change the zoom to 110%, the jitter disappears, but if you scroll the page back and forth you can make the effect reappear.

Thank you! I've been unable to reproduce it under the conditions you describe (or any combination of DPI and zoom). It is interesting that the problem disappears at 110%. 

I expect that your analysis and guess that there is a problem with one of the scripts on the page incorrectly rounding the scrollTop but it's hard to verify without reproducing it.

Oshima, do you have any ideas? Failing further insight I'll file a bug against doubleclick.

elawrence@, can you open https://www.quirksmode.org/m/tests/dpr.html and tell us the value you get for "Reported DPR"?

screen.width equals 2560.
Layout viewport width is 1738.
Expected DPR is 1.472957422324511.
Reported DPR is 1.

Hmm, so it's not high dpi related.

Still unable to reproduce and no further report. Closing for now. Please comment on issue to reopen if it happens again.