Bugzilla

(In reply to Andrew Creskey from comment #4)

> Experiment of increasing the raptor delay between pageloads from 1 second to 10 seconds (android):
> 
> https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=c059749d487e8668fb006b1f247de5f34edc5897&newProject=try&newRevision=20258fcc8dd2d755c82cac08955595552db554dd&framework=10
> 
> I don't know if this would be practical in the lab, but there do appear to be some large perf wins and some regressions.

Your results, combined with the browser settle time paint an interesting picture. For some reason, the settling of browser gives significantly worse test results for some tests like Instagram, but it does reduce the overall noise significantly.

Similar thing repeats with your run with pageload delay modified. Instagram sees a noticeable regression in performance, but noise is reduced significantly. I almost wonder if there is something up with some of the sites.

(In reply to Andrew Creskey from comment #4)

> Experiment of increasing the raptor delay between pageloads from 1 second to 10 seconds (android):
> 
> https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=c059749d487e8668fb006b1f247de5f34edc5897&newProject=try&newRevision=20258fcc8dd2d755c82cac08955595552db554dd&framework=10
> 
> I don't know if this would be practical in the lab, but there do appear to be some large perf wins and some regressions.

Your results, considered alongside the browser settle time paint an interesting picture. 

For some reason, the settling of browser gives significantly worse test results for some tests like Instagram, but it does reduce the overall noise significantly.

Similar thing repeats with your run with pageload delay modified. Instagram sees a noticeable regression in performance, but noise is reduced significantly. I almost wonder if there is something up with some of the sites.