Consistent art would be the biggest, for me personally. The LPC collection is a good example. You can, for the most part, mix and match resources. Also see Buch's stuff, a lot of his tiles would mix and match easily.

I second the complete art packs proposal. Also larger packs you can pick and choose from. Too much content here consists of a single 3D model, or a single sound loop, that are hard to match with anything else (I'm guilty of that too.) Such content can find uses, it's just not easy.

I recently made a visual novel demo using art from here, here and here. (Sorry for the shameless plug, my original announcement got lost.) During the creation process, I discovered that the background set is sorely lacking in interiors, while the portraits are not in fact so suitable for a visual novel, being designed to sit in a frame. So I say what we need is more content that was actually used in a game or tech demo -- anything real -- and as such had the kinks ironed out. The assets I mentioned need little improvement, but that just makes it more obvious that they DO need improvement. It's not their fault, but artists do need a little direction as to what a game developer actually needs in a real game...

* I agree with complete packs. Of course that's quite an undertaking, but it doesn't have to be "complete" - just having a set of art that's consistent (rather than just single one off pieces), and it's then easier to full in the gaps as required. Or maybe artists can look at existing art, and add something to complement it. I don't know if maybe people might think this would be seen as copying/plagiarising - I mean, obviously the art here explicitly allows copying, but people often have an artistic drive to be original, or a fear of being accused of just using someone else's work even if it's legal. But it's worth making the point that here, that really is a good thing for games! FLARE and LPC have shown that there are artists out there willing to create art for an existing set - more of that please :) Perhaps we need another LPC-style competition - or maybe it doesn't have to be a competition with prizes, I think the important thing about LPC was providing an initial set of graphics that artists then built on.

* More character art with animation - static images are fine for scenery, but it's not uncommon to see people release a game character which may look great, but being a single static image makes it unusable in many styles of game (well, they could still be used say as an NPC who just stands there, but not if you want them for a player character or movable enemy). The LPC set, and the FLARE graphics, are great not just for being a consistent set, but also having animation.

* High resolution graphics would also benefit Android, Windows etc. Though I'd say the gap is not just a lack of very high resolution, but a dominance of low resolution graphics. Much of the 2D art seems to have a retro early-90s feel (not that that's a bad thing, but this topic is what there seems to be a lack of), and is below the resolution capabilities of any device today. But having said that, I imagine there's the point that 2D game art becomes harder at higher resolutions (I saw someone making the point that doubling the resolution in each dimension is four times the pixels, and if you're doing pixel art, that matters). In practice my understanding is that most in-game graphics that are higher resolutions tend to be created in 3D, and then pre-rendered if 2D sprites are desired. There are a fair amount of 3D characters on OGA, though often not in a pre-rendered format. Still, in the case of very simple vector art like the balloons example, these restrictions don't apply, so I guess vector art (that can be scaled to any resolution) is something there's not much of.

On the note of 3D models used for rendering sprites, is there demand for taking some of these 3D models, and generating sprite sheets? I'm a programmer but did learn enough Blender to be able to render a static model in isometric (Clint's tutorial at http://clintbellanger.net/articles/isometric_tiles/ was a great help), though animation is currently beyond me. Or has anyone had luck with Blender scripts that claim to automate this? (a quick Google reveals http://wiki.blender.org/index.php/Extensions:2.6/Py/Scripts/Render/Spritify ). I wonder if some programmers skip by the 3D models altogether, in which case generating spritesheets from 3D models (especially those with animations) might be an easy (albeit tedious) way to increase the amount of higher-resolution game-ready art.