Why not cache the HTML/CSS instead of using a screenshot? I'm no expert on web/browser development, so I don't know the difficulty or feasibility in caching just the visible bits leaving a static HTML/CSS snapshot minus any javascript seems, but it seems more ideal IMO since you'd still be able to click links.
Good Q! It is something we are looking at/trying. It's a bit more complex as you have to have a pool of suspended renderers to swap the html into as the user switches tabs (the HTML still needs to be properly sandboxed). Then you include static asset loads, etc it becomes less 'instant'.
Just a thought