My proof of concept pages finally got indexed by google but they only show the xml results in the description, not the rendered HTML (from the XSLT).
It turns out google may report it's self as a mozilla browser but they don't apply XSLT translations, they just present the XML in it's data form.
So the moral of the story is always render the html server side if you think the viewer is a bot (if you use code igniter use their user agent library with is_robot())