Some of the recent posts highlight a big part of the issue i have with the "linking" of roll with failure.
it boils down to this: How something is determined within the game and at the table mechanics-wise does not need to be handled differently depending on what the possible outcomes could be. They really dont. It also causes "issues" and "concerns" even for some of those proposing it.
"getting fewer chances at auto-success" has been mentioned several times - pointing to the issue that lets say "playing the GM" so he gives out as many auto-successes as possible is a "good strategy." if you make a roll you "open the door"... etc. Maybe if its one way you wont even know how confident your character is in the result.
But lets take a different look.
In my games, if you decide to check a shiney sword and a sword with runes on it for magic (by whatever means there is that involves a check for success) then for the player and for the character the process will remain the same. narrate, describe, declare , check, adjudicate etc etc.
There was no "we didn't roll so auto..." there was the exact same process regardless of the underlying truth.
I as Gm don't change my mind because a dice got rolled on whether you can succeed or can fail because the roll is a measure of performance, not a testimony to failure chance existence.
When it is done in my system, your character ends up with a result and a degree of confidence in that result based off how they saw their performance, be it high degree or low degree.
The way i have interpreted the "dont have to roll if impossible" and "dont have to roll if too easy" is by comparing it to old days where that kind of thing was unclear. Every had a Gm have your space marine roll to use a phone book? i have.
those "dont need to roll everything" kinds of rules cropped up a while back and in general started to cover the "just stop making them roll for silly crap."
But somewhere over in indie land of gaming theory they started to morph into a more philosophical mandate - "dont make them roll if..." became "cannot roll unless..." in some peoples minds, alongside the "must have consequences and failure stakes" which at least seems to have not crossed into DND yet officially.
So, yeah, no, i am not going to make you roll for phone books or for tieing your shoes or walking acroiss the street or wiping your bum without falling off the stoop.... but i am also not going to let the rules allowing that exclusion to force a second means of resolution for actions and skills that is dependent not on what is being attempted and how that is done but on the underlying final answer.
The lets go to visit grandma at her house route to grandma's house will be the same, whether or not grandma is home when we get there.
You the player don't have to worry about me changing my decision on auto-success or not, or if i will be "influenced" to make it less likely for you to get auto-success because we will be using the same process either way.
The "best strategy" is not keeping quiet and hoping the Gm will give you more auto-successes but will be to have a good idea, a good plan, a good way to get it done "in character", "in the game" using the story, the scenery, the narrative, your character's strengths and the mechanics to get the best possible odds which do include auto-success and aut-failure where appropriate.
If the result of the de jour definition of roll-to-failure relationship is that "dont ask for rolls cuz if you wait the gm may give it to you without you needing to roll" which really smells to me like "playing the GM" then i consider that a serious freaking flaw and drawback to that whole system, because the players should be , IMO, thinking of the marriage between story-narrative-mechanics and not "what gets the best yields out of the gm".
If someone were to say "hey, dont ask about treasure at the table. Wait. cuz if he does it by email later there is always more." would you be highlighting that as good roleplaying game activity? good "strategy". System working great?
Anyway, thats how i see it.