3e fails at what exactly, because I haven't run across a roleplaying game yet that perfectly simulates anything to everyone's satisfaction. I think it succeeds for many (not all) in giving enough simulation that it is satisfying to many... and really that is all a roleplaying game can strive for.
As far as 4e "succeeding"... again at what exactly? I see this thrown about but really what are it's goals and how are the level of success in which they have been achieved not as subjective or even moreso than 3e?
Here is how it succeeds with me: I have no doubt in my mind that others might dissagree with my opinions, but I'll state them none the less.
First a little background:
I think the fundamental strength of a tabetop RPG over a computerized RPG is that the tabletop RPG is processed and implemented by the human mind. Since the human mind can adapt to just about any situation change it needs to, the game in a sense can do likewise.
Example, in a computerized RPG if I'm fighting an ogre, and decide to say, poke him in the eyes 3 stooges style, unless there is a specific code that says, here is how to poke in the eye, I can't do it. No matter how many times I say "I poke him in the eyes woop wooop wooop!" it won't happen.
But with a Tabletop RPG, processed by the human mind, if I'm a DM and a player says "I poke him in the eyes" I can adapt to the situation even if there is no specific "poke in the eyes" ability.
Similarily if the game has a "poke him in the eyes" attack power, unless the computer has been specifically told DO NOT allow it at X point, the computer will ALWAYS alow me to eye poke.
A Human brain can think, "no that's silly, if you try to poke him in the eyes Y bad thing might happen..."
So in my own opinion, a game needs to concentrate less on the "How to do X" and instead concentrate more on "What happens when someone does X."
And now that's where 4e comes in, because in my opinion, I think it does this
for me far better then 3e ever did.
4e feels like it concentrates on giving me the basics (this is how to attack) without getting overly concerned with trying to give me specific types of "attack" like actions (this is how to attack by poking in the eyes.) The game assumes as a human I can adapt the basic attack to fit the poke attack if need be. In addition it gives me a basic framework to "balance" my adaptations so the results are less unexpected. It also asks me to verify if "poke to the eyes" is a viable option in the first place, rather then trying to anticipate every possible time "poke to the eyes" would be invalid.
What about the powers you ask? Aren't they specific types of attack like actions?
Yes and no. Yes, they are, but in my mind, they're mainly just DM shortcuts. They keep me from having to adapt to every idea a player comes up with on the fly. I have no doubt in my mind, that I could run the game without them just yusing the rules on 42, but the powers give me an "autopilot" in a sense.
It's like another DM saying hey I tested this as a rule, it works well.They in turn can be (and probably should be) adpated to fit the specific situation as warrented. (If for instance there are no eyes to poke.)
Monsters are similar. I don't need the rules to show me HOW the Goblin can shoot a fireball, I already know it can because I decided for some reason it can (with whatever backstory I gave it.) I just need a basic framework for what happens when it DOES shoot that fireball.
These are concepts I had when I started gaming back in BD&D and AD&D times... I always felt the game was giving me basic guidelines, and asking me to modify as I saw fit. (And most of the examples seemed to support that.)
The problems I had with it were mostly centered around it not giving me enough of the basics. (No skills, monsters weren't adaptable enough, classes seemed a bit TOO set in stone, no easy way to feel confident a ruling wasn't way off track damage/challenge wise...)
3e fixed a lot of these issues for me, but because it seemed built in the other way "This is HOW a Goblin can shoot firebals" it opened up a host of new problems for me down the line. It made things more difficult for me, and made things feel more like autopilot was the norm, and not the fallback.
4e seems like a better match for me. It fixes pretty much all the issues I had with earlier editions, but doesn't add the problems I found I had with 3e. So 4e is a more successfull update to D&D for ME then 3e was.
Others may (and probably do) dissagree.