I didn't say it did.
I'm aware of how the mechanic works. But, while it does not change a failed action into a successful one, it does allow you to correct for any failures in planning - failures in planning that are inevitable because the players are encouraged not to plan and introduce the planning aspects of the heist into the fiction through Flashbacks. So for example, you might get to the estate and find it is protected by vicious guard dogs. In a normal game, if you haven't prepared for this eventuality or have no resources to deal with it, you are out of luck. You almost certainly fail as the penalty to your lack of intelligence gathering and scouting of the target. But in Blades, you could then in Flashback attempt to have inserted a supply of drugged dog food through the estates supply chain earlier in the day, playing out the story in a non-linear fashion in response to the challenges you encounter. And if you are successful in that flashback, well the vicious guard dogs are asleep, or if you are less successful you could replace that complication with a new complication if you didn't have an outright success - a suspicious dog keeper with a flashlight for example, or perhaps the kennel master has even called a magistrate out to report the foul play, or whatever.
So I get all the stuff you are saying. I'm afraid you don't really understand the source of my unease with the concepts involved, which among other issues involves statements like this: "A failure on an attempt to sneak may be the GM introducing an patrol approaching that will discover the players, putting them in a bind!" I get that, but now we are not only not interacting with the fiction in a linear manner, but we are also not interacting with a stable fiction. In a traditional RPG, the patrol either exists or it doesn't, irrespective of my failure to sneak. Thus, I could suffer consequences of alerting guards or being caught by a patrol if I fail to sneak, but I could not summon up a patrol or create more guards by failing to sneak. Remember what I said about my preference being "fiction first: mechanics second". To introduce a new fictional element on the basis of a metagame construct inherently means mechanics are driving fiction and not the other way around.
In any event, I don't attend to derail this thread into a conversation over the benefits and limitations of a Nar based approach to a Heist game, or to rile up BitD's passionate supporters.