If you have never had to contend with a problematic player, count yourself lucky. This isn't some theoretical problem: real people will sit at your table and ruin the experience for everyone else involved for their own pleasure if you let them.
To me, this is not a GMing problem. It's a social problem.
I don't need, or expect, the maker of a DVD player to tell me what to do if an unpleasant person turns up to my video night and starts spoiling the movies for everyone else. Similarly, I don't think the rules of an RPG provide a solution for people who want to spoil the game for others. It's a social problem, which has to be solved socially.
If the rules of the game have a tendency to generate problems, though, then that's a different matter. Suppose, for instance, that the rules present choice A and choice B as equally viable. (To give an example of what I mean by this, the 4e PHB doesn't suggest that any particular factor is relevant to choosing a class other than the flavour of character that you want to play. It doesn't say, for instance, that shy people might prefer to play an archer ranger rather than a warlord, or that people who find it hard to choose from options might want to avoid a wizard and a polearm fighter.) And then suppose that, in play, it turns out that those choices differ in some respect that the rulebook didn't flag: for instance, it turns out that option A tends to generate twice the spotlight time for that player as option B, even though both players wanted and were expecting the same spoltlight time.
In such circumstances, I don't think it's any sort of solution for the game designer to just throw it all on the GM and say "Hey, it's your job to sort this out." I mean, what does "sorting this out" even mean? Should the GM just block half of the first player's action declarations in order to halve his/her spotlight time?
Nor do I think it is helpful for the rulebook, or the gamer culture more generally, to label players who have chosen option A as "munchkins" or "problem players". (There is a fair bit of this in the 2nd ed AD&D PHB.) If option A is an option that is bad for the game, then why is the rulebook permitting it?
Instead of rather abstract admonitions to "put the game first and participants last", I find it more helpful if a rulebook actually discusses the expectations around the various options and gives advice for handling them. (For instance, Burning Wheel expressly gives the GM and other players veto rights over certain PC build choices. And it expressly flags certain other choice with a warning to players that taking that choice could make for a PC who may have a hard time prospering in the game. Those warnings and permission don't make the
social problems go away - eg someone makes one of those choices and then whinges about what happens to his/her PC in the course of play - but at least they go some way to flagging parts of the game system which are "use at your own risk".)
What Gygax is talking would very much apply to a player in my secondary group who just last week, I kid you not, spent a good 6-8 hours doing NOTHING but looking through D&D 3.5 material to try and find feats that would boost his animal companion (he's playing a druid). He actually had the guts to straight up ask the GM, "I think you should let my wolf take Feat X" (I can't remember what it was actually called) "which lets me use him as a riding animal."
So, he wanted his 1st level druid animal companion--an average, normal wolf--to take a feat designed for ACTUAL riding animals . . . just because the player wanted to. And to top it off, the player's character is a full-sized human. Not a halfling, not a dwarf . . . a human.
Of course, after the GM read the feat in detail, it was totally against RAW (as written the feat would have required the character to spend a YEAR training the animal, even if it was the right size to be a riding animal, which it wasn't). Should the GM just have allowed it? I don't think so. And more to the point, if you know this player, you'd know it wasn't an expression of a deep-seated roleplaying desire, it's one of pure power gamesmanship.
I guess my issue with this is that I don't see how Gygax's quote applies. I mean, this is a player asking "Do you mind if I cheat?" and the GM answers "Yes I do. Please don't." I'm having trouble seeing how there is much more to it than that.
That said, maybe I can see a little bit more: the greater the extent to which a game relies upon the mechanical rationing of fiddly little bits for the purposes of PC building (and 3E/PF would seem to be the pinnacle of this), the greater the likelihood of a conflict between a player's conception of his/her PC, and what the rules actually permit him/her to build. And houseruling in a solution becomes tricky if it is not clear whether the mechanical limitations serve some sort of rational balancing purpose, or are simply there to enforce someone's conception of verisimilitude, or are just there because someone thought it was a good idea.
To give examples: the level-limits on the two-weapon fighting feats seem intended to achieve a balancing purpose, by interacting in some appropriate fashion with the stepping up of iterative attacks via BAB. (Whether or not they succeed at this I leave as a debate for another thread.) At the other end of things, the alignment restrictions on paladins and monks, in 3E, are there simply because the 3E designers thought it was a good idea to preserve certain traditions. It was stated back at the time of 3E's release that they serve no balancing function.
With the particular issue you are referring to, I can't tell whether the purpose of restricting the feat to certain sorts of animals is to do with balance (it would be broken for a 1st level druid's animal companion to have that feat) or whether it was simply to give affect to someone's idea of verisimilitude (humans can't ride wolves because they're too big). In accusing the player of trying to cheat I'm assuming there is a balance issue, but if I'm wrong about that, and it was just a verisimilitude thing, then I feel a little sorry for the player. If there's no balance issue with being a wolf-riding human druid, then it seems a shame that the game rules get in the way of that.
Labelling of the reasons for pre-requisites could help here (are they for balance, for verisimilitude, or "just because"?). In the absence of such labelling, I'm not sure that Gygax's advice helps much because how can a GM tell whether allowing the option in violation of the rules as written will hurt the game (eg multiple second-weapon attacks at 1st level) or not (ignoring the alignment restriction for paladins and monks)?