The bolded bit of your post is the explanation of the bolded bit of @Manbearcat's that you have said is wrong.The bolded section is just wrong. I have no idea what you mean by "when a conflict is over" without the framework of a skill challenge or timer. Whatever is happening is happening, and if a player wants a particular outcome, they have a bunch of actions available to change the world to their liking. A conflict ends when the PC has gotten the thing they wanted, or it is no longer possible for them to get the thing they wanted.
This is the thing I keep talking about, where the PC sets the victory condition, and gets to change it whenever they want. They decide what it is they want, and then spend their available actions to get to that point. The story of a given campaign is the ex post facto stringing together of that PC chasing each of those goals and recording what happened from point A to point B along the way.
More to the point though, removing the GM from resolution is the primary goal of such a system. Actions have discreet effects and consequences, so you can figure out what happens by plugging a PC choice into the mechanics for resolving that action, and be handed a result. Then you reevaluate the state of the world, any other parties that can take actions do so, and you repeat. Continue until everyone is dead or the PC has gotten whatever it is they wanted done.
Or in other words: in the approach you are setting out, it is the GM who decides the elements of the fiction, the causal relationships between those elements, the "state of the world" after each action is declared and resolved.
Return to the castle infiltration example:
A PC bribes an off-duty guard to learn the password: Does the guard know the pass word? Has the password changed since the guard was last briefed?
The PC dresses up as a guard: How strict is uniform compliance in this castle? What accents are regarded as foreign or otherwise suspicious?
The PC enters the castle: Is there someone in the castle who recognises the PC from when they were shopping at the market two days ago without a disguise?
Meanwhile, another PC scales the castle wall on the "lightly guarded because impregnable side": Is a guard looking in their direction? If they throw a rock over the wall to make a noise and distract the guard, does it land on someone's head, thus raising the alarm rather than causing a distractions? When the PC wants to drop down into the courtyard having scaled the wall, is there anyone there who might notice them?
The PC dresses up as a guard: How strict is uniform compliance in this castle? What accents are regarded as foreign or otherwise suspicious?
The PC enters the castle: Is there someone in the castle who recognises the PC from when they were shopping at the market two days ago without a disguise?
Meanwhile, another PC scales the castle wall on the "lightly guarded because impregnable side": Is a guard looking in their direction? If they throw a rock over the wall to make a noise and distract the guard, does it land on someone's head, thus raising the alarm rather than causing a distractions? When the PC wants to drop down into the courtyard having scaled the wall, is there anyone there who might notice them?
On the approach you're describing, the GM has to make decisions to every one of those questions, and each one of those decisions has a big effect on the players' chances of having their PCs succeed. On a skill challenge approach, the skill challenge resolution settles them all. (And similarly in other closed-scene resolution frameworks.)