Fixing the core chassis would be the first thing. Of course there's so many things needing fixing this is much easier said than done.
Off the top of my head: Most classes are frontloaded, poorly conceived based on ideas that hold little to no weight in a real game*, or otherwise critically flawed in one way or another. Fighters are more akin to templates than classes, paladins are only a slight improvement, still failing conceptually and mechanically for the most part, prepared vs spontaneous casting is way off, many feats are on par with skill tricks, if that and certainly aren't worth an actual feat... I could go on for several pages but you get the idea.
Now how to fix that? ToB was a good idea. It makes martial characters what they should have been from the get go. Even if not applied in that exact form, giving such characters more options in combat and actual class features *cough*fighter*cough* is a big improvement towards fun, or at least avoiding monotony. Against myths they aren't actually stronger, a full attacking barbarian is doing at least as much damage as the warblade, for example. But the warblade has more than a small handful of battle options. That's a start. Now let's look at casters. A cleric, without supplements has little trouble being a better martial combatant than the martial combatants and still has a good deal of their spell power for divine wrath. This is just horribly wrong. Of course, they deliberately made clerics overpowered so people would play the 'heal bot', something that if you're resourceful does not even apply. Druids are actually worse. Two words. Natural spell. Ready made super gish, anyone? If there's one thing druids do right though, it's that it's a class worth taking 1-20. This is true of few others, very few others indeed. Though, it does so for the wrong reasons (anything else would just weaken it), the idea is right.
In my eyes, the sign of a solid, balanced class is one where you can feel taking it as a career is a viable option 1-20 - you need not be required to multiclass or prestige class to maintain effectiveness, though you can choose to do so if you wish, and this is neither too weak nor too strong either. And although they won't do this, it should require as few books as possible to accomplish your goal. Since they make more money if they make fewer books, they won't. But at the very least no class should require material from any book it isn't published in with the exception of the basic ones to be workable. As an option? Sure. Never a requirement. But as again they make more money with more book sales, they really don't have any incentive to not make more 'fighters', aka wait a while while we pump out extra material slowly for you classes.
I am well aware this is far easier said than done. One of the key things though is playtesting - get it right this time!* Want to learn what needs to be fixed? Recruit people determined to break it, let em go at it and see what happens. I might be thinking of something else here, but I believe they did get the Character Optimization board there involved, which was the best move they could have made. If they can't collectively break it, it can't be broken.
* - The game is balanced around the core iconic party, built with elite array and played by tactically inept players. I've seen games with one, or two of the above but never all three. If nothing else, characters will be played somewhat effectively if not intelligently. Yes, the 6 int half orc barb can deduce how to smash things properly. Is smashing a good idea at all times? Not really.