EDIT: I kinda got lost in my rant and realized just now I kinda reiterated, if not repeated, almost everything Cry said. O_O But my personal story might be of some use here!
I figure since Cry wrote a lengthy post with his opinions, I should contribute as well!
I do agree with you on the idea that the word 'demo' is used cavalierly. I do agree that the best beta testers are, well...beta testers. Not the entire modding community. I do also agree that when demos don't receive amazing feedback they fizzle and die.
I don't know if you've read about it, but before OLitD I was building a story named The Scarecrow. This as a while back now, when I was less experienced and, quite frankly, very naive. I was trying to build a quality custom story while also being a full-time student and an avid video gamer. Basically, I never left my dorm except for class XD
Anyways, I released a 'demo' of my story, which was the product of quite a while, probably something like 6 or 7 weeks of work. It wasn't bad work, but most of it was just a showcase of atmosphere and some neat scare events. I got some healthy feedback, but soon after I kind of lost steam. I'd spend significantly less time working on it, especially on weekends. I remember I would sit down for 3 hours building or detailing a level, but once I didn't get the huuuge flood of approving comments I was expecting, I kinda kicked back and just worked when things got really dull.
So I can attest that if you hope to make a quality product, you had best keep things on the DL until you can release a polished version of your work. If you release a subpar or buggy demo, interest in your project may fizzle and so might your motivation. Another mistake, I believe, is to release a demo which is basically all work you have made up to that time. It seemed to leave a huge tear in my story. I had thought "Ok, the demo is released, so that part is all done. Now I can move on to the next part and completely ignore the last". Unfortunately a lot of the work I had worked on after the demo (but before my unfortunate hardware problem) really didn't fit in with the demo, and I would constantly be going back and tweaking things on both ends...kind of like trying to tape a torn piece of paper back together, but when the tear still shows and you rip off the tape, you mess up the paper even more.
Now, when you release a demo, you want to make sure that the end of the demo makes sense, and matches up with the storyline. A lot of demos, as Cry said, simply end with you walking through a door or falling over and seeing a grunt walk up to you. This, in addition to being overused, makes no effing sense. Your demo is, presumably, out there to spark interest in your project. How do you spark interest? Cliffhangers, hooks, unique features and storyline, etc. Make your demo have a climactic, even episodic ending; something that will leave the player saying "NO WAIT I HAVE TO SEE WHAT HAPPENS NEXT!" instead of "Oh, crap there's a grunt. I'm dead." or "Oh look a new level. I wonder if I'll trip on my way in and pass out, waking up in the full release."
Anyways, the whole spiel I'm trying to give here is that a demo should indeed be a furnished product, just like your full release!