There are two kinds of game balance with regard to power level. First, the party's power level versus everything else, usually a dungeon or series of antagonists. Second, the party members' power levels versus each other --not necessarily in a PvP sense, but in the sense that they are consistently able to out-perform the rest of the party.
Temporary Superiority
Sometimes a party member is more suited to a particular situation, which is fine as long as it isn't a deliberate choice on the DM's part. If you feel like using undead, and the cleric is better at fighting undead, it's all right for that character to exhibit a superior ability to deal with his area of specialty.
Other times, a character is better at a particular level, but this is temporary. Fighters are (much) better than wizards at first level, in vanilla 3e D&D. This goes away as the characters increase in level, and is in many ways fun -- the wizard "pays" for his later superiority by being weaker early on. Indeed, he "earns" his later power merely by surviving.
The danger here is if there is a major trend in superior level. Even if it's just that the player is better than optimizing his build, it's a problem. Indeed, it's probably a bad thing that a character design system can be so well gamed, that players who are not as interested can be left so far behind in balance.
Lowering Top-Tier Effectiveness
In Team Fortress 2, a first-person shooter game that came out a few years ago, the game designers made the conscious choice to limit how much more effective a skilled player could be. Shooting someone in the head doesn't always inflict more damage, for example. Nor does having played longer give you access to weapons that are per se better, though it does give you more (sometimes more newbie-friendly) options.
The result is that, even though a skilled player is still obviously better during gameplayer, it isn't an automatic shut out. New players with quick reflexes can still compete with the pros, or at least contribute to the performance of the team. There is a narrower band within which player skill can improve that player's performance.
D&D should be the same way. More proficient (or interested) players should have an advantage for seeing rules possibilities that other players don't, but not a huge one. Playing D&D should be more about tactical decisions than character design. This doesn't sell rules manuals that are essentially catalogues that players can peruse through to "shop" for bonuses, but it does make for a more rewarding play experience. The player should be most rewarded for decisions made on the fly, not what happens when he's picking feats before the game begins.
When It Goes Right
Ideally, the adventuring party should be a smooth functioning machine, with each party member performing his equally important role under pressure, in dangerous situations with outcomes that are only partially predictable.
When things are working appropriately, there is a rising arc of action, with smaller encounters in a dungeon sapping some of the party's reserves and testing their ability to allocate resources, with mild possibility of things backfiring horribly.
Then you fight the boss (or whatever), and the boss truly tests the party's mettle. It's the apex of the rising action, making a player death or even a wipeout more likely than in the normal course of things. Great stuff.
When It Goes Wrong
Nullifying a particular player's advantages is a mistake. If you start deliberately crafting encounters that are resistant to an above-average power level player-character's abilities, or that hit on all his weaknesses, he will usually feel like you're picking on him. Using any form of negative reinforcement will just make that player unhappy.
The way to fix things is to increase the power level of all the party members, through means that bring them specifically into line with the other player. For my game, I introduce new spells for the underpowered casters' spell lists, extra class-specific feat options for under-powered rogues and fighters, and magical items that are more likely to help the under-powered players than the over-powered fella.
Does this make the entire party too powerful? Absolutely. But monsters and traps can always have their difficulty ramped up. Have them do more damage, have a higher armor class and saves, and generally amp things up. It might be a little faster than you expected, but that's fine, it's all in the game. And it's certainly better than having an incredibly over-powered player making everybody else feel like schlubs.
