There’s plenty of failure in Mouse Guard, even in missions that Luke runs.
I knew you were going to say that 
Anyway, I took Luke’s comment to heart and it improved my games immensely. Here’s how I figure it.
If I sit down to take an exam and I answer the questions wrong, I have failed. I wanted to pass, but I failed. I suck.
If I sit down to take an exam and a fire alarm rings, or my rival has switched up the answers, or I’m attacked by an alien, I haven’t passed the exam, but I haven’t failed either. I have not sucked.
There are a lot of Twists in a Mouse Guard session. If I repeatedly describe your mouse sucking, even if things keep moving forward, you may leave the table feeling like your mouse sucks. I’ve seen it.
In Mouse Guard, I don’t describe mice failing when I throw in a Twist. If you’re sneaking up on a badger, my Twist is not that you step on a twig and the badger guard is roused. You’re sneaking beautifully, when something happens…things go wrong and you may indeed end up in a fight for you life with the badger, but it’s not because you failed to sneak well. That seems to work for me. It’s sometimes a challenge, but makes for a much better experience even if the player never, ever achieves their goal.
When I run Burning Wheel, it’s different. I’ll happily describe how you misstep while sneaking and jump into a battle with the guard.
Edit: I’m not trying to say this is what Luke meant. He is perfectly able to say what he meant. This is what I do.
James, I understand what you’re talking about, but that’s a separate issue.
Try this: if you get a condition, then whatever situation you were dealing with was resolved favorably.
If you get a twist, then whatever situation you were dealing with either remains unresolved, or has been resolved unfavorably.
Thor, what are your thoughts on giving the player a choice at that point (after they fail the test)? Ponder the options and then speak out loud. Make the consequences explicit and metagame a little. So they could either gain (specifically identified) condition, and resolve the situation favourably or encounter (specifically identified twist) and the situation remains unresolved or is resolved not in your favour. Choose.
I’m in favor of the GM making the call. Delivering a good twist is the best part of being a Mouse Guard (or Torchbearer) GM. Why give that away?
What I like about the MG approach is that you can have plenty of failure on a lower level, but generally not on the higher level.
You can fail to overcome an obstacle, or have a “scene” go to hell, but the rules make sure that the game never grinds to a halt (like in “you missed the vital clue, you lose, end of the game, suckers!”). A mission may fail, but the twist keeps things going, because now the twist becomes the important thing. And if the GM cannot think of any more good twists or wants to go to the next turn, he can always hand success for the current important thing to the players by imposing conditions.
Also, if a hero never fails, that’s a bit boring.
Yes!
Sure. Once the Twist is done, the story goes in the direction the GM decides fits the developing fiction. If the crow flew away, the crow is gone.