For (1), it seems that the first thing the GM needs to do is to vary scene-framing imperatives.
Here's Edwards on something in the neighbourhood of what you're describing - both possibilities and risks/pitfalls:
Simulationist play works as an underpinning to Narrativist play, insofar as bits or sub-scenes of play can shift into extensive set-up or reinforcers for upcoming Bang-oriented moments. It differs from the Explorative chassis for Narrativist play, even an extensive one, in that one really has to stop addressing Premise and focus on in-game causality per se. Such scenes or details can take on an interest of their own, as with the many pages describing military hardware in a Tom Clancy novel. It's a bit risky, as one can attract (e.g.) hardware-nuts who care very little for Premise as well as Premise-nuts who get bored by one too many hardware-pages, and end up pleasing neither enough to attract them further.
Historically, this approach has been poorly implemented in role-playing texts, which swing into Simulationist phrasing extremely easily, for the reasons I describe in "
Simulationism: the Right to Dream". You cannot get emergent Narrativist play specifically through putting more and more effort into perfecting the Simulationism (which requires that the Narrativism cease), no matter how "genre-faithful" or "character-faithful" it may be. I consider most efforts in this direction to become reasonably successful High-Concept Simulationism with a strong slant toward Situation, mainly useful for enjoyable pastiche but not particularly for Narrativist play at all. . . .
Sole reliance on deepening and detailing any aspects of Exploration is misguided. The vast majority of attempted Narrativist design is a hunt for the perfect Simulationist design that will ostensibly permit the Narrativist play to emerge, leading to abashedness at best. It's often combined with mistaking an effectiveness-improvement mechanic for a reward system - at this point, the game text simply facilitates High-Concept Simulationist play, and the Narrativist goal is left to Social Contract alone. Various publishing practices, especially a long string of scenario and setting supplememnts, provide the coffin nails.
With the above in mind, one challenge I can see is this: how do the players know what sort of scene is being presented to them? One possible answer: what's presented in the scene might be sufficient to let the players know whether it's intended to be an exploration scene or a protagonistic scene.
I think the bigger challenge than this scene-framing aspect is on the resolution side. Upthread I quoted Edwards, from "setting dissection", saying "embrace the fullest and most extreme rules-driven consequences of every single resolved conflict, no matter what they are" and to "Show those consequences and treat them as the material of the moment in the very next scenes, every time". But that might not work for the exploration scenes. The issue here, it seems to me, is how to make sure the ingame-causality that is used in these scenes doesn't undercut the elements of the shared fiction that are central to the protagonistic moments of play. In the abstract we can talk about the importance of "quarantining" one from the other, but how are we going to do that if the same PCs are in the same setting interacting with the same NPCs? I'm not saying it can't be done, but it seems tricky to me.
As for (2), it looks like it might just be another version of (1), or else it might be "story now" with very extended framing.
This remark seems apposite, though not conclusive, in relation to (2):
"El Dorado" was coined by Paul Czege to indicate the impossibility of a 1:1 Simulationist:Narrativist blend, although the term was appropriated by others for the blend itself, as a desirable goal. I think some people who claim to desire such a goal in play are simply looking for Narrativism with a very strong Explorative chassis, and that the goal is not elusive at all. Such "Vanilla Narrativism" is very easy and straightforward. The key to finding it is to stop reinforcing Simulationist approaches to play. Many role-players, identified by Jesse Burneko as "Simulationist-by-habit," exhaust themselves by seeking El Dorado, racing ever faster and farther, when all they have to do is stop running, turn around, and find Vanilla Narrativism right in their grasp.
Maybe group (2) is looking for a very strong explorative chassis, manifested via what I've called "extended framing". The challenge is how to stop it turning into those Clancy-esque hardware pages? And how to avoid that without play just being agonisingly slow? The problem is that GM narration on its own isn't RPGing, but every time the players declare actions that are resolved via considerations of ingame causality rather than protagonism we're back to reinforcing simulationism and running the risk of undercutting protagonism. I don't know if there's any general solution to this. As Edwards says (same essay):
What happens when Premise is addressed sporadically, or develops so slowly that the majority of play is like those hardware-pages? Whether this is "slow Narrativism" or "S-N-S" or just plain dysfunctional play is a matter of specific instances, I think.