Fumbles. They've never been anything more than an optional rule. While there have been, on occasion in different versions of the game, printed rules for fumbles, these have never been part of the core rules. Yet it seems most groups use them. That's not a bad thing.
However, I have come to the realization that most GMs use fumbles without considering the full implications of the set of rules they are using. So I'm going to lay out the implications of fumble rules.
1. Fumble rules always hurt PCs more than they hurt monsters. This is because a PC makes thousands of attack rolls over the course of a campaign, whereas a monster that lasts 10 rounds and full attacks each time with 3 attacks (total of 30 attacks in his lifetime) is considered an exceptionally lucky beast.
2. Further, fumbles also hurt PCs more than monsters statistically. D&D combat favors the PCs. This is because, if the monsters were always equal in power to the PCs, the average campaign would last no more than two combats, since each combat would have about a 50% chance of ending in a wipe. Anything that increases randomness favors the underdogs; thus, fumbles favor the monsters.
3. Fumbles render mass combat entirely unrealistic. If 5% of attacks end in Something Awful happening, a battle between armies of more than 500 degenerates into a flurry of dropped weapons, snapped bowstrings, and people flinging their greatswords into their friends' backs. How does one fight a war in a situation like that? And, if that situation's not there, then how come the PC Thog, the Master Barbarian of the Crescent Serpent, with the skill at arms to defeat 90% of the world's blademasters, is flinging HIS sword into his friend's backs 5% of the time?
4. Fumbles hurt warriors more than mages. This is actually my sticking point on them - melee characters in D&D, aside from a few exploitive builds (infinite-damage hulking hurler, invincible frenzied berserker), are already struggling to keep up with the mages in terms of effectiveness. Why saddle them with even more trouble?
5. More skilled warriors fumble more often than less-skilled warriors. This one is entirely counterintuitive and counterrealistic. If a natural 1 either fumbles or threatens a fumble, then a character with +20 BAB (4 attacks, not counting any extras from two weapon fighting, Haste, a speed weapon, or other such abilities) is going to, round by round, roll four times as many fumbles as a level 1 commoner with a stick. He's swinging four times as many times; he's going to fumble four times as many times. In games that used particularly brutal fumble systems, I've found myself voluntarily forgoing my lower iterative attacks - a choice that costs a melee character a lot of his power.
6. Critical hits already have a counterbalance. Further, that counterbalance already occurs, automatically, on a natural 1 attack roll. That counterbalance is called "missing." A critical hit, unless you have an additional ability to make it better, deals, in most cases, less than twice the damage a normal hit does - bonus dice, which nearly everyone who makes attack rolls regularly accumulates over their career from class abilties, feats, and weapon special abilities, are not multiplied in a critical hit. Thus, a fighter who rolls a critical hit and later in the combat rolls a miss has effectively dealt damage as if he was constantly rolling normal hits.