As it happens, there is no unwittingness here. Fail forward as I understand it from the infinite arguments of the mages in this forum, is where we get success-with-complication. No fail and find another way. 5th edition isn't fail forward, that's just one option within the 5th edition consequences-resolution system.
Fail and die, for example, is not failing forward, but it is consequential.
You've stated 3 outcomes:
- you open the safe (the consequence you want)
- you open the safe but with additional consequences
- you become engaged with some consequences
This is classic "no dead ends" or "no whiffing" or "something interesting always happens and the situation always changes." That is Fail Forward resolution. It undergirds the Burning Wheel family of games (from BW to MG to TB), D&D 4e, 13th Age, among others. PBtA games make use of this on knowledge/perception/divination related moves that result in a 6- (yeah, you see the thing or know the thing or perceive the thing...but I've got bad news for ya).
As a resolution technique, its designed into games for three purposes:
1) So the dead end of "no, and the gamestate/fiction remains unchanged (nothing happens)" is taken entirely out of play.
2) To prevent deprotagonization due to a high "whiff factor" which screws up expected levels of competency (Control-F "Whiff Factor") and reduces player control of the gamestate/over the momentum/trajectory of play.
3) In concert with a trivial genre credibility test (eg no jumping to the moon without escape velocity boots) + intent-directed conflict resolution (eg not "can I open the safe" but "are the documents in the safe no strings-attached"), it gives players huge authority over the momentum and trajectory of play because it basically takes GM veto off the table.
However, every game that I know of that uses it has codified DCs and they're table-facing so GM mediation within the architecture of action resolution is virtually nil. A game that takes that off the table is bringing in GM veto or control over the momentum/trajectory of play by proxy of establishing the DC. Further still, a game that doesn't encode the results of "yes" but allows/requires the GM to interpret that also brings about GM control over the momentum/trajectory of play by proxy of the "there is still work left to be done" button.
Further, my understanding (and application) of 5e Ability Checks is that the GM is in control of everything beyond the player's action declaration. Order of Operations for GM:
1) Impossible to succeed "No"? Impossible to fail - "Yes"? Neither and meaningful consequence of failure (MCoF) - "Maybe"?
2) If MCoF set DC based on either genre emulation (presumably based off of the natural language interpretation of an adventurer within the current Tier of Play; "what tropes can heroes/adventurers of this Tier of play - distinct genre - pull off?") or process logic/internal causality of the world (presumably off of a natural language interpretation of and orientation to Very Easy, Easy, Moderate, Hard, Very Hard, Nearly Impossible by "a character with a 10 in the associated ability and no proficiency").
3) If failure, then the GM can decide retroactively to use Success at a Cost (fail by 1 or 2 means success but some kind of cost or complication akin to the 7-9 PBtA Defy Danger result or a new obstacle/setback akin to Twist + Fun Once in TB or 4e's Skill Challenge micro-failure) or Degrees of Failure (5 or less might mean "nothing happens/gamestate is unchanged" or "gamestate changed somewhat adversely but not catastrophically" while 5 or more is a "botch" with serious gamestate consequences).
4) If Crit Failure then calamity might ensue (at GM's discretion) or Crit Success then maybe an extra boon (at GM's discretion).
5) Decide how much the gamestate has moved/how much of the situation is left to resolve (presumably based off of the GM's conception of the convergence of (i) what makes for "fun" + (ii) "what makes for a compelling story" + (iii) "what tailoring to this particular player or this group of players demands" OR (iv) some kind of prior precedence for consistency of handling when the demands of consistency are at odds with (i-iii) ).
That doesn't look like what you've laid out above (which, again, looks like Fail Forward). "No and/or the gamestate doesn't move (which can be interpreted as "meaningful consequences for failure" and clearly is by a cross-section of the user-base...typically with evinced Gamism proclivities or "we're there" experiential proclivities because people fail and nothing happens within the setting their adventure occupies)" is absolutely on the table in 5e. And Genre Emulation based DCs and process-based/internal causality (based off of the perspective of an in-setting "character with a 10 in the associated ability and no proficiency) DCs are both the domain of 5e DC setting (and they will be different because they're different baselines for Easy, Moderate, Hard et al).
Whatever else 5e Ability Check handling is it is profoundly GM-directed in a myriad of ways. Its not Fail Forward. Its certainly not Gamism. Its certainly not Story Now. Again, this is why I've been calling it a devise for GM-Directed, High Concept Simulation with possibly a veneer of Process Simulation. It doesn't do nearly enough work to get to actual Process Simulation, but a GM might mix in a nice chunk of both approaches to DC handling in 2 that a player might feel that the experiential quality of play hews enough to respect for internal causality that their Process-Sim Dander doesn't get kicked up. Exceeeeeeeeeeeeeeeeeeeeeeeeept...this is a big time violation of "the consistency clause." So I don't know how GM's resolve this. My guess is that the overwhelming % of 5e tables are composed of and tailored toward players that don't particularly care about "the consistency clause" and they just want a fun, relatively casual time, with a tailored experience of compelling story + performative theatrics and plenty of color + a tailored experience that hits the Power Fantasy notes they're looking for. Whereas a table for
@Crimson Longinus and or
@Thomas Shey better have a GM that consistently hits the internal causality grounded litmus test for DCs and action resolution.