We had one of our engineers programming our HD2 back when it was still in testing. He ran it as a free-form rock format. It ran for several months and then we shut it down and handed it over to programming who decided to wait on it until they could come up with a strategy (we weren't too bothered as it was mostly a test anyways and we hardly expected it to live forever). However, I still get occasional calls or emails from GaTech kids who want it to come back!
HD2 and HD3 can be viable if wisely programmed. It's not going to make you a ton of money, but you can probably cover your expenses and if you have HD on you are essentially paying for the bandwidth anyways, may as well use it. The only trouble is that wisely programming it would require an investment of time and research or at least some old-school experimentation by people in programming, and given the climate of the day (especially with the big radio groups) no body wants to do ANYTHING unless they are guaranteed success.
One of the big troubles with HD from a programming standpoint is that a lot of people still don't really know what it is and what to do with it: Should sub-channels be treated as complimentary programming to your main (niche formats and the such), or should they be treated as independent units? Should they be advertised on your main frequency, or is that cheating on yourself? Is simulcasting a sister station a viable option (great way to get some fill-in on a weak AM station), or is that a cop-out? Should it be treated like block programming, or should there be a 24-7 format? These are some of the questions that really need answers. The best way to avoid a stumble is to know what the terrain is first.
I don't think HD is a non-starter... I do think it has some more growing pains to go through before it proves successful (and it will probably never be as successful as analog was, but then again today's analog stations aren't that successful either). I do agree that the technology and deployment could definitely have been done better.