What are you feelings regarding the current trend of making failure "attractive" as an option by virtue of "failing forward" in game design? To what extent does this impact the legitimacy of a challenge? If "the game must go on!" eliminates some failure scenarios because the players might find those scenarios undesirable...would you say that that undermines the concept of failure? Should failure from a PC stand-point be undesirable from a player standpoint with undesirable consequences that accompany it?
Good questions, and I will do my best to answer them.
It is my belief that the GM should not "guide" the action to a specific outcome. On the other hand, I do believe that part of good design is seeding a location with enough material to allow unexpected things to occur, for both good and ill. It is my firm belief that the early TSR designers, for instance, did not expect every last bent copper piece to be found, and thus seeded treasure in excess of what would usually be found so that, if a player happened to think to look inside the giant lizard's gullet, there was a chance of actually finding something.
I am well aware that there are some who imagine that every scrap of treasure in a published scenario is intended to be found, even though the only quote available on the subject, in Module B1, says exactly the opposite: "Although monsters will inevitably make their presence known, treasures are usually not obvious. It is up to players to locate them by telling the DM how their characters will conduct any attempted search, and it is quite conceivable that they could totally miss seeing a treasure which is hidden or concealed. In fact, any good dungeon will have undiscovered treasures in areas that have been explored by the players, simply because it is impossible to expect that they will find every one of them." (p. 24).
There is nothing inherently wrong with offering the chance to "fail forward" in a scenario. Having unexpected good come out of failure can actually offer sweet moments in the game...but the word "unexpected" is key. Like the expectation that finding cool treasures spawned the idea that they would follow you around until you did discover them, the idea that some form of good could come out of failure ceases to become surprising when there is reason to expect that most failures will be anything but that...failures.
What JRRT called "eucatastrophe" - the feeling that, when catastrophe is assured, sudden hope changes everything - is a powerful feeling, but it only works when catastrophe seems inevitable. It doesn't work when failure is expected to be "failing forward".
From the previous series, it should be clear that I think that the maximum good, from a player's standpoint, comes from being able to play the associative game. The sort of metagaming that comes about from deciding what forms of failure are off the table works against this. It also works against the idea that the player's choices have any value - if the choices lead to failure, so be it. I have written in the past that the GM should never include a consequence for failure that he is unwilling to live with, and that is because players must be allowed to fail if they are ever to experience true success.