Wishlist for 4E:
1) Greater modularity of rules components, and less interdependencies between them. D&D gamers are rules tinkerers, and a game designed with the ability to rip out large chunks of the system to suit their needs is a hoop I'd like to see 4E jump through. You can, for instance, rip skills, feats and AoO out of 3E, but it leaves game balance wounds (because the game assumes that a fighter's feats are balanced with a rogue's skills, etc.) and a lot of annoying interdependent bits and pieces to ignore. For 4E to keep each layer of the ruleset as a self-contained module would be a game design feat in and of itself (pun intended) but that's the kind of excellence you'd need to aspire to to make a new edition worth it. Having said that, I doubt it's even truly achievable...
2) Relating to that, a cutdown on the statting overhead of the system. It's almost okay as is at low and medium levels, but there is still definite room for improvement, especially at high levels. One or more parts of the system might have to go on a diet to achieve this. If 3E is really designed to be fudged on the fly, make the system reflect that in more ways than just token ways such as a generic NPC stat list.
3) Fixing of the skill system so that the DRs tend to break into always succeed/no chance to succeed a bit less at high level. Ability checks seem primitive and blunt tools by comparison, but at least their resolution was less likely to be a foregone conclusion, and less completely unpredictable from party to party at mid-to-high level.
4) Greater building on the fluffy-crunch intellectual property of D&D from editions past. Although designers are justified in attempting to present something new for the new edition with regard to fluffycrunch material, the time for that comes when the crowning achievements of past editions are already in the game, not vice versa. Unless this is some keep-backwards-compatibility-low-so-they-buy-new-stuff conspiracy on behalf of WotC (which I doubt), leaving out death knights in place of digesters for game design reasons is keeping too much of an eye on the game balance ball at the expense of the game's flavour and feel.