And if the players feel that the DM is cheating and setting arbitrary numbers designed to thwart their strengths, they may well get offended by that.
The DM cannot "cheat."
He can run a good game or a poor game. Sometimes running a good game means thwarting the players. Sometimes running a good game means letting the players succeed. For some players, exactly the converse is true; and for some players, they don't want the DM to make those judgment calls at all, but to simply let things fall as they may.
Point is, with such a micro-managed system as 3e, a lot of player choice has specific, defined, expected advantages within gameplay that is thwarted if you "cheat" on your monster/NPC creation and just do whatever makes it as hard as you think it should be for the PCs. You are short-changing the choices they made, and the knowledge they've gained and put to use about how the game is expected to work.
If the strength of a system is meaningful choices, you undermine that when you make those choices meaningless.
That's a fair point, but it
is predicated on the assumption that "The Game" is a
competition of mechanics, and the story is only a framework designed to showcase those mechanics; as opposed to The Game being a cooperative story, and the mechanics merely facilitate the unfolding plot twists so that the outcome is unknown by any of the participants.
Certainly there's a sliding scale. Some players define "meaningful choices" as "I will take this feat in order to eke out a +4 advantage to hit," as opposed to, "It might be distasteful, but we'll have to ally ourselves with the lich in order to take out the hobgoblin army."
In the grand scheme of things I am less dissatisfied when a monster proves too difficult to hit based on the
mechanical choices that I made, as opposed to the dissatisfaction I feel when the
story choices that I make are thwarted (ie, railroading).
I have found that the most satisfying games balance the mechanical choices to just the right ratio of success to failure to keep the story outcomes in doubt.
This is not an edition-specific observation. All sorts of D&D games have been run up and down that scale through all editions.
Now, as an edition specific observation, 4e has rather famously balanced the success/failure ratio to an optimal point; and in order to discourage "unwanted behavior" on the part of players, they have stripped away many of the "meaningful choices" that "mechanics" players exploit: Such undesired behavior is disincentivized by explicit codification into the rules.
Some players need that because they simply won't behave otherwise. If you want to run a game that focuses on the story, but you are constantly thwarted by players who define "win" in mechanical terms, you have a problem (a problem that 4e solves in a particular way).