Go to a site that is offlined by a service worker. For example, my blog: https://blog.wanderview.com Open the network panel and reload a few times. Often we finish loading in 200ms to 300ms, which seems reasonable. Sometimes, however, we hit load jank which puts us at 1 to 2 seconds. This might be partly due to sending all our resources in large chunks instead of properly streaming. I've been told we can trigger main thread jank when trying to parse large single chunks of html, etc at a single time. We should check this again once bug 1204254 is fixed.

I don't think this is an e10s specific issue. While testing e10s vs non-e10s, though, I can only reproduce it in a browser with many other tabs. Its probably just cross-tab stuff blocking runnables needed for load on the main thread. I'm going to leave this open, though, to retest when I streamify things.