Skinning is generally poorly supported in MediaWiki; there's lots of stuff that should be under the control of the skin but is not (e.g. thumbnail framing). It would be nice to get that (more) right with MCR.
Combining content should depend on the role (and maybe the page type), but also the skin. I like the idea of getOutputPlacementHints for dealing with positioning the content within the skin, but the role handler needs to wrap the content HTML, and there should be a way for the skin to replace that with its own wrapping mechanism, IMO.