It's been answered several times, but I feel the need to respond anyway. Sue me.
1) A model that is buildable. For me, this translates to a model which is at least capable of being assembled into something which resembles the intended subject, and that means sometimes I don't feel the need to correct models I unfold from computer games. They are just too close to being ready as-is to be worth hunting for every intersection. Sometimes this doesn't work, but most of the time I get complimented on my models because I choose ones which I know can be built with a minimum of work reshaping parts.
2) This greatly depends on the person, but I've found that between 8 and 18 inches to be pretty commonly accepted. Some go for larger, and some prefer smaller, but this is where the middle of the curve generally lies. The reason is that this is that 12 inches is a common shelf width, and if you angle an 18 inch model it can fit on the shelf while still having a decent profile.
3) PDFs, PDOs, and image files, in that order. Scalability is the primary reason I hear. PDFs are slightly more popular because they don't require you to download special software. Do NOT stave in JPEG, though! It is notorious for losing color information, which muddies the image and makes text harder to read. Use PNG instead. It takes more memory, but it is a lossless format.
4) Anything, really. Nobody has complained yet about my 19 MB Moldy Crow file.
5) It depends on how intuitive your model is. Usually, Pepakura is sufficient, because it can point out where each piece goes and how it's unfolded. If your using PDFs, then a set of instructions goes a long way. Just including some step-by-step pictures is fine. The Grissom model by Martin Sanger had only pictures to go by, and they were fine for the task. Just an exploded diagram was included in the Bird of Prey on Billy's Star Trek site, and I had no problem assembling that, either. Always try to beta-test your models, though. Have someone build them and send you notes and corrections before you put the model out there. It is totally worth it, though not everyone does it. I don't usually, because I have a high 3D perception and can "build" a model by sight. You might be able to do this, too. But at least consider doing it. I have had some of my more complicated models beta-tested, and they turned out better for it.
Bottom line: there will always be complainers. You can please some people some of the time, but you can never please everybody every time. So don't try. Deal with the ones with questions, but follow your gut and learn from experience. If someone complains, respond with, "You don't like it? Make your own."