Lacking the right software is not a valid excuse. UT will send "ALA requests" every once in a while if the lending library isn't in their software. As far as I can tell an ALA request is just this form mailed or emailed: http://www.ala.org/rusa/sites/ala.org.rusa/files/content/sec...
I might suggest contacting various universities about getting copies of theses from them. It's possible that they'd be perfectly willing to scan them for you, even for free. On this note, I've been surprised by the extent some corporations have gone to provide me with copies of proprietary technical reports they produced. One time I called the number on the website of a large corporation and my call was forwarded to their staff librarian, as I recall. A few weeks later I got the requested report. It had to go through some release process, but they were happy to share their work. Another time I emailed a generic address at a Shell research lab, and a few weeks later I got a copy of the report I wanted. There have been failures as well, but I was surprised by how frequent the successes were.
Snobbery is a pretty good reason for anything, I've found. In any case, it could be many things: expense (as my university regularly reminded us, each ILL cost like $20 on net), low demand from patrons (even if self-fulfilling, still a valid reason), lack of trust, not being plugged into the right ILL system/software... I don't recall any of the books I ILLed from my public library being indicated as being from universities, though there were several close to us.
I suppose I should try my public library when I have some spare time. I'm almost certain they'd be unable to get either books or theses or papers, but it'd be interesting to know the specific reason why not.
> There are many paid document delivery services. Here's my university's one:
Yes, actually, I know someone who used that one last month. (Scan quality could've been much better, IMO.) It is, unfortunately, rare to have a straightforward 'pay us $X and we'll give you a copy of any thesis' service linked or mentioned on the library website, and they are only for that library's holdings ("scans of book chapters and articles from the UT Libraries collection" ie not anyone else's). I would complain a lot less if most universities had it! I've sometimes wondered if more university libraries have it than I think they do, and they just hide it.
> As for embargoed theses, my impression is that usually the author wants to get a patent on something discussed in the thesis.
That would be reasonable, but in the subjects I usually research, that would make little sense. I think the last embargoed student thesis I ran into was a study on the stimulant effect of nicotine on cognitive performance; hard to see any patent on that being possible, much less profitable to apply to.