I confess to having read parts of this thread rather quickly, as I'm still catching up on mail after a month-and-a-half-long absense. But I did notice one thing: generally, "TEI" and "XML" have been used interchangeably.
The two are not equivalent; TEI is an initiative that has guidelines, and is not a markup language. Nor is "semantic encoding" (which is at least partly what we're really talking about when we discuss digital editions) equivalent to "XML." It is quite possible to create, say, a digital edition of a literary work without looking at the TEI Guidelines at all. In order to allow it to "talk" with other editions, and with some emerging tools, the creators might want to then translate it -- not "into TEI" but into another set of elements and attributes, one that is in accordance with the TEI Guidelines.
TEI has been absolutely essential in promoting dialogue and framing our thinking about digitized texts, no doubt. But it is not essential for doing humanities computing, or for encoding, just like a food processor is not essential for making bread.
I'm not qualified to speak to other issues raised in the thread (such as the relative value of facsimile and/or encoded digital editions); just wanted to point out what I see as a fairly major element in understanding humanities computing.
-Vika