Making a low res sprite takes a few seconds, making a 3d model with good textures takes a day of work.
First of all, I'm not equating HD with 3D. I'm suggesting that 2D games should be made with high-resolution art, ala Braid, Earthworm Jim HD, BlazBlue or King of Fighters XII. Essentially, I'm talking about using a 128x128 character instead of a 16x16 character.
Also, ambiguous graphics work in much the same way as an abstract messy brushed painting, your mind fills in the details.
While that may be true, It's... conveniently true, and not really applicable to the medium. I'd like to use an example above and compare two graphics from an old, 16-bit game and it's modern, HD counterpart:
<img src="http://img833.imageshack.us/img833/6396/earthwormjimremakescree.jpg">
Apologies that it gets a little cut off in the forums, open it up in a new tab to view it in full. What you should notice from left to right is that the character is drawn in the same stylized way he was drawn in the first game, just in higher fidelity thanks to a higher-resolution display. The level design aesthetic is largely retained as well.
In his book Understanding Comics, Scott McCloud addresses the same property of art you're talking about: that stylization allows a person to project onto the character because the character is intentionally vague. This is what allows for comics like Sin Titulo. But resolution of image is a system limitation, not a stylistic choice: it is an artistic challenge, not an artistic medium.
Take this image from the first page of Sin Titulo, downgraded to approximately old-school resolution:
<img src="http://img254.imageshack.us/img254/2724/sampledowngrade.png">
You cannot possible argue that it is "just as good" as the original despite the resolution change. It isn't. Details are lost, expression is lost, the text isn't readable and has to be replaced with a font that doesn't fit as well, et. al.
The underlying fact of the matter is that low-resolution art resources are easy to create. It is easier to program an engine for low-resolution resources because sloppy programming will not be as noticeable in the frame rate. It is easier to script for low-resolution art because it's easier to predict motion with less pixels on screen. In short, low-res is easy?but again I say, it isn't a stylistic choice.
There's a reason Megaman 9 and 10 were sold at bargain bin prices on Xbox Live Arcade despite being manufactured by a major industry giant: they were retro, requiring little effort or money to make. They recognize that the graphic choice tweaks our nostalgia-bone, and that the gameplay is still fun. But they also recognize that you can't possibly market a never-made game from 15 years ago as worth the same money as a modern game. After all, I dare say most of the users on this board could create a new Megaman game that is faithful to the originals using Construct. Had they made new, HD Megaman titles, they would have sold them as disc titles like they've been doing with the Megaman X series, if only because the initial investment would have been higher.
Ultimately it comes down to this: it is more difficult to tell the quality difference between Gods and Super Metroid than it is to tell the difference between This Game is Hard and Braid. We are no longer in an era where sellable games can afford to have their art created by programmers instead of artists, because even Joe Shmuck can tell when the art is poorly made.
Look, for example, at Cave Story and Spelunky: both games are either available or "coming soon" to consoles, and both featured low-resolution art. Notice I say "featured": now that they've got some funding, both are getting a graphical overhaul. This isn't a coincidence. Low-resolution art is sellable only for nostalgia value: it isn't now, it isn't valuable.