Hehe...Well, ideally they'd finish every game, but I think in general it's much, much better if a game can really get you hooked right away or at least after a half hour or so. I know some games don't get interesting until hours and hours of gameplay, and that's a bit much. Same with books or movies. If it takes until you're almost done with the book before it "gets good," that's a bad book in my opinion. That said, I have no problem with a game developer who expects you to read a manual or do something else to prepare before getting in. It may be more popular to cater to the lowest common denominator, but I respect games that don't.
I agree 100% with the idea that you should finish things before moving on. If I were serious about making a career at game design, I would take on projects and see them through. The only game I've made is a text adventure using only C++. I learned a lot--and I mean *a lot*--that I never knew before about coding and how games work just from that, and it didn't even have graphics. I can only imagine how much more you could learn with a more ambitious project.
I have also had the pleasure of talking to several "1/2 man band developers" who *were* making a living at it (not to contradict your general point, of course). One of them was making upwards of 50,000 a year just making one casual game per year. His strategy was to target game portals and market his games to "non-traditional gamers," such as women and older people. His games were very polished and slick remakes of things like Qix. Sure, he wasn't a millionaire, but he was thriving in his dream job without having anyone over his head. From what I hear, working for most game companies (EA spouse, hrm) is just undoable for family men.
Also--I've heard the same about the game design doc as well. Many people have "good ideas" but they're not really fleshed out. Sitting down to do a detailed design doc can really help them flesh out the idea and anticipate all kinds of problems, not to mention giving them a much better idea about what resources to bring to bear. With so much great middleware out there, I don't see why one man or a small team couldn't bring a game to market. I see it all the time in the casual and adventure game sectors.