It's not the players who search and do or don't find the mace - it's the PCs.
The reason I emphasise that distinction is because it relates to the other distinction I've been making - between the ingame consistency or "reality" of the gameworld, and the real-world activity of authoring the gameworld. Within the gameworld - from the ingame perspective - it is not "Schroedinger's mace". It was never there, as it had been taken from the ruins by the dark elf.
But at the table, the authorship is in response to the check.
This also illustrates an aspect of "fail forward" that has been discussed upthread, that what is at stake is not just (or even primarily) whether or not the PCs succeed in their task, but whether things turn out as they wanted.
This difference from "fail forward" highlights what, to me, is key - that "fail forward" is a technique that is intended to drive dramatic and narrative momentum without GM pre-authorship. (At least as it has been systematised in those games which make a point of calling it out as a technique.)
I see your distinction between the two, but for me I like there to be less of a distinction between What the character does and what is authored at the table. As a player I want to "look through" the PC's eyes. I want to decide what to do based on what he or she would do, not based on what would make a "better story".
I see why you like the avoiding of "pre-authorship", as it is something you dislike in your games (possibly as strongly as others like having it). For you it is a negative, for others a positive. I like the style of game that results from this kind of play and the benefits of the more "advanced" fail forward techniques don't really add enough to the game for me to compensate for what I feel they take away.
As a player I like to have as similar perspective as my character as is possible. That is why I like checks to align closer to what is being checked for.
I can see why in you example of the mace you are happy with the mace never having been there. I see how it works for your game and your style.
For me the check has a real concrete meaning (I know it doesn't for you). It means they are searching for something. The check determines if they find it or not. but that is where the 2 approaches differ. I explained my preference above, and I see why you like your preference. The reason it grinds against me is because the roll is being modified by characters abilities, to determine something totally unrelated to the character. Why does bill the bumbling idiot trying to search result in a 80% chance that the mace is on the other side to the continent, but if Omar the Observant looks there is 20% chance.
For me that is not a fair use of probability.
I would much rather know the DM was playing "fair", and knew where the mace was, and the search result told me how good my searching was.
Once the GM knows what the answer is to the location of the mace in advance, and that is taken as fixed regardless of the dynamics of play when (if at all) the search actually occurs, then the focus of play has moved away from dramatic momentum to something else. (Eg exploring the GM's world/mystery/etc).
How does exploring the DM's world not have "dramatic momentum"? How is the DM knowing where something is supposedly detracting from the game? My games have a lot of dramatic momentum even without using more extreme versions of fail forward. The DM could well know that the mace is there or not there, so the same possible outcomes are still all there. The difference being only did the players roll determine the location and the finding or just the finding?
In the discussion of the trapdoor begun by @
Nagol upthread, the possibility of a jailer turning up to taunt (and perhaps free) the trapped PC was discussed. If that jailer is authored by the GM not in advance, but because it is a useful element of the fiction for maintaining narrative momentum, then the logic is no different from "fail forward" as illustrated in the example of the mace.
But whereas in some approaches to GMing needing to introduce the jailer post hoc is seen as a failing that, ideally at least, would be avoided, in "fail forward" games that sort of approach is made integral to the running of the game.
Too far in respect of . . . ?
The adding of a jailer is an interesting question. I have no problem with the DM adding a jailer to what is happening, it is interesting and keeps the game moving forward. I don't see a point in adding a jailer only if you roll low on an ability or skill check. If it is interesting to add a jailer then add a jailer. adding a jailer only if they roll low on an unrelated skill doesn't add anything to the game as far as I can see. If the DM sees that the action is slowing or the narrative is not moving and so decides to add something to the what is happening, why tie it to any skill roll?
My preference is that complications get added and characters use their skills to deal with those complications, rather than skills are assumed to work and failure adds complications. After all complications will happen to both good and bad doctors, but the good doctor use their skills to deal with the complication. ( I realise it is a different way to look at skills than you are used to).
If fail forward is seen as a series of approaches (many of which I use) this is going to far from how I want the game to run. I am happy using grades of success and failure, alternative fail conditions such as time or creating too much noise, letting rolls stand, etc.
For me the cut off point is when the PC view and the player view separate so much and when skill checks determine unrelated effects. (starting of rain, Schrödinger's situations.) I clearly see why they are not a problem for you and that they can enhance the game for you and others. For me they don't work for the games of D and D that I want to have. (ironically I think they work fantastically for leverage, and think it is a good fit).