We have a concept where we test the second page view (not repeat view) where we access one page and then measure the next page we visit. It's a important metric since we then can see how the browser cache/local storage will affect metrics.
We have had problems with it before when WebPageTest report too fast metrics and we have that again.
The TTFB is too fast 0.036s and the start render is completely wrong: 0.119s. Checking the filmstrip view we can see that start render happens after 2 seconds. Checkout this example:
http://wpt.wmftest.org/result/160312_53_4S/
We had the problem before with the WPT register start render when the Chrome login popup changed but that is fixed. Need to check again what's wrong, I though did a fix where we make the screen white before we do the second page view (to make the filmstrip view and video cleaner) but that seems to be gone now because the filmstrip view starts with the first page view.
