What an interesting question!
That happens to be almost exactly what I have been puzzling over for a couple of years and getting nowhere so far, though I have a provisional workaround.
In my case I have a store on hard disk containing many .htm and .mhtml files (saves of part/whole webpages) that I want to scan using a two-pane window. The LHS pane would give some kind of hierarchical navigation tree by category, and the RHS pane would display just the headings and/or first line of each page in that category, and be able to expand into the full page on demand. A bit like my bazqux feed-reader.
