I agree that designing the game around bad GMs isn' the way to go. Not really sure where he is going with this though.
I think I understand this.
Have you ever met someone that spoils an otherwise good or acceptable situation for an entire group? Usually, it's because whoever is in charge of that group doesn't feel like they can directly address the person that is being the problem, they feel that they have to make a ham-fisted, blanket policy to neutralize the situation rather than deal with the one person that's being a jerk.
In an office with a company I once worked at, we had an intern that showed up late. The intern was talked to about it and the office manager considered the situation resolved. The intern kept showing up late, habitually... So rather than deal with that intern, the office manager enacted a new scheduling policy. He started emailing people their schedules every week, and people in the office were required to print out a copy of the schedule, sign it, and include it with their end-of-week paperwork. An entire office of forty people was required to do extra paperwork because of a
single intern. To no one's surprise, the intern kept showing up late. The office manager wanted to make a policy requiring everyone in the office to take a picture of the posted schedule with their smart phone and set the photo of the schedule as their phone's background. Fortunately, the intern left the company the week that policy was to go in to effect and everything went back to normal.
That office manager is exactly what I thought of when reading about trying to get the game's design to insulate the game from "bad" DMs. I don't understand why the office manager just didn't get rid of that intern. In the same vein, I don't understand why players stick with games they don't enjoy.
In the game design sense, I think that the column is pretty clear. I've had experiences with the sort of DMs that make you functionally guess until you figured out the solution--particularly some DMs that were being "clever" by using medieval riddles in everything and making you "roleplay" the solution. Of course, the "solution" is always a single approach that's the "right" solution, and they shoot down other ideas.
Player: "We've been trying to figure this riddle out for two hours. The heck with it, I'm going to kick the door in."
DM: "What, are you just going to give up? Why don't you try roleplaying to solve the riddle?"
Player: "I don't know the answer and you won't allow me to make any sort of check to see if my character does. I want to do something else today besides stare at the miniatures clustered around this door. I'll kick it in."
DM: "You try to kick the door in and your foot bounces off. It's banded with adamantine and you can tell that it's reinforced from the other side. You know that you won't be able to kick it in."
Player: "Fine. What are the walls made of?"
DM: "Uh, stone, I guess."
Player: "I take out my pickaxe and sledgehammer and chop through the stone wall."
DM: "You chip away about an inch of stone and find a thick sheet of adamantine in the wall. You can't make a hole around the door. You have to guess the riddle and open the door to get in."
(Yes, this is paraphrased from an early 3E game that I played.)
In this sense, I think that what Mike Mearls means is that the game shouldn't flat-out tell the DM that they are "not allowed" to make an impenetrable adamantine vault in an adventure designed for 3rd-level characters that can only be entered by figuring out an obtuse 11th-century French riddle. Depending on inclination and ability of the DM, a skill challenge or even the rules design of adamantine can be interpreted as binders or straightjackets on a DM's options for how the players are challenged, both mechanically and narratively.
Rather, the DMG (or equivalent) should advise the DM to make several possible solutions to any given challenge and not to force the characters (or their players) to "guess" at what they have to do. The rules shouldn't necessarily force a DM to use a skill challenge for any challenge that's not a combat encounter, for example.