I thought the point of this thread was for clutter discussion specifically, not orbiter design discussion... but ok.
Thom, it might even be easier that way, since the "old Orbiter" could still exist as the "new Orbiter" is developed. People could choose which to use - stability or new functionality - on a per-device basis if it were set up that way.
The key points I'm interested in (no design stuff in the database, re-design-ability without any code changes, and platform portability) have been addressed, and I'm very happy with that direction so far.
There are a few other items I'm interested in - the ability to create your own custom screen(s) on a per-installation or per-device basis, better media sorting, and user management.
For example - having the "climate" and "security" rows are useless to me. I would rather make a screen with a clock, some media shortcuts, and controls to set lighting levels as the home screen. The design I would want to do would be selecting what I want, and where to put it on the screen - the complex stuff like picking icons and colors should be handled by the orbiter. This way if you change the "theme", or whatever you want to call it, my custom screen will still match everything else. Maybe this complete customization is only available for 1 or 2 screens, where the rest are dictated by the design that is shared by all users.
For another example - say I have a device that doesn't quite fit the mold of any other - maybe it's a cable box that has arrows and select buttons and stuff, maybe it's a PS3 which has 51 (I think) buttons on the remote (many of which have nothing to do with media and will exist on no other device), or maybe it's some completely different device that has nothing to do with anything. It would be great if part of the device template could describe what buttons should be on the remote and have the orbiter pull the proper things together automatically. Think of a generic remote template - where buttons or groups of buttons appear, disappear, or reorganize themselves based on the device's capabilities. Of course, part of the UI design will describe how to lay out the buttons and what icons to associate with them - the template only knows what buttons are important. Maybe there is a priority scale for selecting which buttons to leave out if the screen is too small.
For media sorting, this should be completely customizable. The user should be able to make up hierarchies - like Author, Album, Title or Title, Season, Episode - and attach those to buttons on the orbiter. The same should be true of any sort of search - for instance "TV Shows", "Action Movies", "James Bond Movies", "TV Shows recorded in the last week", "Unwatched Videos", "Recently watched videos", "Videos partially watched", etc. Maybe even shortcuts to specific playlists. The user can select which ones are important or make up their own, and they appear on the orbiter. I know exactly what I usually search for - there are only 3-4 searches that I do - so I should be able to make shortcuts to those searches so I don't have to wait for the datagrid, select the search buttons, run the search, browse through the results because the search is only close to what I wanted, and finally make a selection.
Media sorting should also be configurable for any search. For instance, it's easier to look a James Bond movies by release date, videos should be sorted alphabetically (leaving out the "The", as I've stated before...), and recorded TV should be sorted newest-first with new episodes (this week's new stuff) highlighted. These are my preferences - other people will want other sorting, I'm sure.
The thought behind these ideas is that no one likes the same things, and no one has the same technology implemented, so why should everyone use the same exact interface? Slight customization in some key places can bring the UI to a whole new level very quickly.
I've also been wondering for a while if there is a better way to handle the current user of an orbiter in the case where no one is using it. Currently, someone is always the user of an orbiter, but as new technology comes about where we can figure out who is in a room and react to it - that assumption stops making sense. This brings up the case where no one is in a room, and the case where more than one person is in the room. Orbiter can't handle either of those, currently.
These are probably much further down the road, but something to think about, if you haven't already.