It was confusing. These are the simplest changes. Any rule that commonly resulted in player or DM confusion--whether due to unnecessary complexity, unclear wording, or simple errors in design--tended to draw attention from the developers. A lot of the changes in the combat chapter of the PH fall into this category.

It was too good, or not good enough. Third Edition had a lot of playtesting before publication, but it had a thousand times more since its release in 2000. That's bound to find the "weak points" in a game system, whether those are spells that are too good (haste and harm come to mind) or classes that are too weak (such as the bard or the ranger). It's OK for some options to be more useful than others (for instance, fireball is more often useful than dispel magic), as long as the choice of "what's more useful" can still vary dramatically by situation. If characters make the same choices over and over again, that points to the likelihood that those options are simply better than the others. If every sorcerer begins every fight with haste, that option's probably too good. If every specialist wizard chooses the same banned schools, maybe those schools aren't good enough. If no spellcaster ever bothers to carry around a magic staff--an enormously iconic magic item if ever there was one--then the staffs in the game need to get better.

It didn't fit the spirit of the Third Edition D&D rules. Third Edition introduced a lot of sweeping changes to the way D&D works, much of those coming from a goal of streamlining and templating the game systems. For example, attacks, saves, and checks all work in the same manner now--you don't use a different rule for making an attack roll or making a skill check. Where we found places that the rules didn't meet that goal, we tried to adjust the rules so that they did. Monsters in 3.0 work much more like characters than they did in previous editions (they have ability scores, for instance), but there were still too many places (such as skill and feat acquisition) where they didn't. The methods of giving out skill points and feats to monsters were all across the board, and none worked exactly like the method for characters. Version 3.5 codifies all that into a single unified system, which is much more in keeping with the way Third Edition D&D works. The "square" facing of all creatures in D&D is another example--rectangular facing for big creatures establishes a default "facing" for the creature, something that the 3.0 combat rules tried to eliminate. These rules weren't necessarily confusing or unbalancing, but we believed that revising them would ultimately prove advantageous for players and DMs of the game.

It didn't work correctly. This is a tricky category of changes, because it encompasses a variety of issues. A spell's description might be worded in such a way as to confuse players as to its actual effect. A feat might be least useful to exactly the characters who are most drawn to it. A monster might be so complicated to run that DMs couldn't use it effectively. None of these rules are necessarily broken, but none of them work in the manner they were intended. These aren't "necessary" changes in the way the first two categories might be, but D&D shouldn't be a bait-and-switch game. If Power Attack looks most exciting to the greataxe-wielding barbarian (and the game sells it that way), then he should be the one getting the biggest bang for his buck from the feat, and not the twin-dagger-wielding halfling rogue. The game should deliver what it promises.