The roll was pointless if they can just retry, because at that point success is guaranteed. That's why I spent half of my response to Thomas Shey talking about gambling for progress. And how if it's just a matter of rerolling until you win, just skip the rolls and move on because the outcome is predetermined. This is why traditional games, largely, require very similar responses to rolls, by the DM, as PbtA games enforce structurally. Because having no response, likely, causes the outcome to be predetermined.
This is also my response to Lanefan here;
But he clarifies;
So he's largely doing as I was intending. By limiting retries he gives the roll some weight. There is a real meaning to a failure assuming success had a real meaning to begin with.
Now obviously, if the required change is cosmetic and meaningless, the roll can still be retried ad nauseam, and the initial problem still exists. But if the change required to retry costs something, then not allowing retries is sufficient to give the roll meaning. The roll changes the status quo.
This is why rolling on a chance of success, without consequence, is a problem. And why PbtA-style reactions to failures are needed in traditional games, albeit in a slightly different form. Because the outcome becomes predetermined if you don't.
So you have to add cost to the failure of a roll, or you simply are prompting pointless rolls. The FrozenNorth post I cited demonstrates what happens when a roll has no meaningful consequence. The roll can be repeated until the players get the desired outcome, and that's a waste of time because success is guaranteed.
I hope that helps clarify what I meant. I think the response applies to Lanefan too, which is why I used his post as an example.
TLDR: Not allowing retries can often be sufficient to solve the problem I am alluding to in my posts.
EDIT: Hawkeye did a better job explaining my position than I did. Hmph!