I’m not exactly a 4e hater. (Even though laser clerics and astral diamonds and starblah armours are profoundly stupid. In my opinion.)
So, there exists a bunch of things that 4e does very right. I have not played it, so these are only from design perspective. (My bias: I like elegant game designs.) The following are not in any particular order.
Out of combat
Skill challenges are a development long overdue. They allow one to mechanically handle non-combat encounters in such a way that it takes some time, which focuses more attention on them. The challenge can be constructed so that it promotes using different skills, or at least accepts such use. The difficulty can be scaled arbitrarily by increasing DCs or the number of successful checks one needs to achieve a victory. (Two ways of handling difficulty are redundant, as one would have been enough, but it is easy enough to always look difficulty from a chart and only mess with the number of successes required. Or the other way around.) Skill challenges allow partial successes, which are essentially a form of “Yes, but…”. You track down the beast, but it has time to slay the residents of a lone farmstead. You find it resting atop a heap of slaughtered farmers. Good luck you did not fail two checks or it would have ambushed you. Or three, because it would have lead you to an ambush by an unfortunate band of orcs and slipped away in the fray.
Skill challenges are not actually mechanically interesting. To make them gameable, one would need to leave hints about the applicable skills in any particular situation.
The challenges are modular; if you don’t want to focus on a particular thing, just call for normal skill check and be done with it. Unfortunately the gamedoes not allow one to do this with combats, as of yet.
There are less options at character generation, and radically less options when advancing a character (no multiclassing). Both of these make the relevant process faster, which is good, but reduce playable options, which may be bad. The supplement treadmill is likely to greatly increase the number of options, given time (and money or illegal downloads).
At higher levels when getting a new level one does not so much gain new powers as swap old ones for new ones. This is good, because it reduces the number of options one has in play, hence reducing analysis paralysis and makes it less likely that some ability is forgotten (I have lost a 3rd edition character because I forgot he had one fifth chance of negating critical hits, and it was not fun to remember it afterwards). Also, swapping powers means that planning the character’s path 19/29 levels into future is less necessary, though not any less rewarding, which I think is a good thing.
Non-options, like 3rd edition caster/different caster or caster/noncaster multiclassing, have been radically cut down. This reduces the role of system mastery in character generation, which I think is a good thing. Nonfunctional archetypes are no fun.
Rituals deserve their own entry. Personally, I think that rules which force one to make choices between combat and noncombat ability are a bad thing in a combat-centric game. For example: Preparing fireball or whatever third level utility spells there exist in 3rd edition. This is not a problem in games that do not focus on combat to such a degree. Actually, the problems mostly arise in games where combat encounter, as opposed to say an entire dungeon, is a discreet and central unit of game. Utility spells do not always or usually function within that unit, hence it makes sense to make them a separate resource.
The idea of rituals also fits my aesthetic preferences. Implementation not quite as well.
All characters have several, hopefully viable, actions to take during any given round. At least in theory. This is certainly an improvement from 3rd edition, where all characters have a number of theoretically viable but often practically useless options. And then there is grapple.
I am certainly intrigued by how well the roles and their special abilities actually function in actual play. Does the fighter pushing a target by one square as an at-will power actually make a difference? This I’d like to know.
Here’s a bit of game design philosophy I support: Rules are bad if they are not used in actual play. Hence, the simplified monster stats are, in my opinion, a good thing. They reduce unnecessary cruft from the rules.