I posited this idea because when Rule Zero + hidden backstory are both in effect, ultimately it is the GM's job to eliminate standstill. The players can only guess for so long at the hidden backstory elements that are preventing them from achieving their goals.
A Rule Zero GM must at some point break the barrier by either providing new vectors of information, or reframing scenes in a way that make new player action declarations possible.
This is interestingly no different than the stated GM agenda for PbtA play --- "adhere to your GM principles and make your moves".
The difference is that PbtA provides additional player side authority to largely eliminate stalemate play from appearing in the first place.
I would say a principal difference from AW is that, in AW, (i) "nothing happens" is not a GM move, and (ii) the GM moves are all framed by reference to player aspirations for their PCs ("put someone in a spot", "announce badness", "offer an opportunity, perhaps with a cost", "separate them" (when they want to be together), etc).
So it's not just that the GM eliminates standstill; at every point, the GM is inciting the players to declare actions for their PCs by putting the players' aspirations for their PCs under pressure.
In the sort of "rule zero" play you're describing, either the GM does the same thing, and play drifts to (what The Forge calls) "vanilla narrativism" - that's what happened with me in the second half of the 1980s - or else the GM puts out "hooks" which the players are expected to pick up on, and then play drifts towards something fairly trad.
A table that doesn't want either of the two options described above therefore needs to abandon "it's the GM's job to eliminate standstill". This is the approach taken by classic dungeon-crawling.
But in Rule Zero play, it's all on the GM. (S)he is fully responsible for developing valid game states that grant the ability for players to make action declarations.
If the goal is to avoid stalemate / "rowboat world" play, a Rule Zero game must rely on the GM.
I don't think this last-quoted sentence covers the field: see my previous sentence just above. I think classic dungeon-crawling puts it on the
players to eliminate standstill. But this requires some fairly tight conventions on how scenes are framed, and how certain canonical actions (involving doors, ropes, 10' poles, etc) are resolved: we can see those conventions presupposed and occasionally stated in Gygax's AD&D rulebooks and in Moldvay Basic. I do think that one problem that can affect "sandbox"-type play is a lack of conventions, analogous to those which apply in the dungeon-crawling context, that govern scene-framing and some fairly canonical action declarations.
I think one of the reasons sandbox play has largely never taken over as the primary mode of play (vs. trad) is that GMs regularly fail to recognize this responsibility. If you take away player inputs to change the fiction, then some other inputs must fill in.
My diagnosis is a bit different. I think that there is significant degree of reluctance to acknowledge that classic dungeon-crawling depends upon
conventions to make it viable: and hence there is a failure to develop comparable conventions to govern play that wants to use the same sort of GM prep + player resolution of stalemate as dungeon-crawling does, but in a more expansive fictional setting than an austere and artificial Gygaxian dungeon.
An interesting feature of Torchbearer is that it does set out conventions that govern the full gamut of play. It doesn't just say "make it verisimilitudinous" and then punt everything to the GM.