In my tests, using content.scrollTo is much faster at rendering content than updating the displayport. The problem is, that we see gray lines (see bug 612599).
Can this be fixed now? And would it have any impact on getting "scroll" events when content scrolls (e.g. bug 654122)?
Updating the scroll offset is no longer faster than updating the displayport thanks to incremental displayport updating, and it doesn't always work if the user pans to somewhere that doesn't change the scroll offset at all. As we discussed on IRC, scroll events will currently not happen if the page dimensions are not big enough to extend the CSS viewport because the scroll range is 0, 0. I think this is the right thing to do, but we have floated the idea around of allowing "bogus" scroll offsets that won't make sense from a content POV. At any rate, I think we should probably close this as WONTFIX. If we'd like fake scroll offsets, let's file something else. Objections?