Different artforms have different expectations, and some of the games already residing under the ever-stretching aegis of "RPG" are so different from each other as to be effectively different artforms. (Fuckin' Yay!, I am all for ludological diversity. I just don't give a shit about the state lexicon. It doesn't change fast enough.)
But that's not to say it's impossible to model one artform within another, or to build a third thing which has aspects of both. I'm not very interested in classifying today's or yesterday's games into this camp or that camp. I'm more interested in tomorrow's games, and I think the imaginary "divide" between "narrativist games" and "traditional games" is more an accident of genealogy than a valid categorical distinction. But the problem domain is broad and complex, and our discussion gets confounded when people use words like "story," "plot," and "narrative" interchangeably. This has added much mud to the water over the years, leading some (including me) to either devise their own jargon, or appropriate words from other fields or define them in special ways.
The main effect has been that game design discussions that broach tender subjects - not unlike political discussions - often tend to bifurcate into two polar opposites, even when it makes no sense to think of them as such. These tender subjects directly include obvious things like "plot," "arc," or "railroading," and less obvious ones such as what others have termed "Creative Agenda" and "Shared Imaginary Space."
I try to remain mindful of this, and yet my design interests perpetually move me directly into these regions. I find myself bridging some sort of gap, or trying to make others see the bridge I'm standing on. I'm a hybrideer, a biplaystylist. Given this desire for biplaystylality, what can be said about Plot that everyone can agree on?
Let's see if I can offer something to start with.
First off, we all must admit something: If a system's primary priorities are to be plot-based, then clearly other considerations will have to be descended in priority, at least to the degree that they must be safeguarded from undermining the primary priorities. Therefore, subsystems for guidance of those secondary priorities will need to be established. This is a structural issue that impacts not only the Player's immersion and the quality of the emergent narrative, but the very flow of functionality. It is important. It's even more important than Player expectations. Here's why:
1. Imagine a modern branching drama game (each scene with a fixed number of possible "exits" or "paths" to other scenes - these paths may be visible like "cutaway scenes" in videogames, or just assumed and hidden between the cuts. In this game, the Player expectation is that the system provides a dramatic arc, a consistent world, a coherent plot, a cathartic climax, and a karmically-fitting denouement, like a well-made movie. I don't necessarily need to explain all the kinds of choices you (as a Player) might make before playing, because (for example) I can build a system that prompts/queries you as you go along, whenever you hit a node (i.e. a branchpoint requiring resolution).
2. Ok, now imagine another game with the same plot-based expectations, but where I don't build paths between the nodes at all, I just scatter proactive NPCs, reactive objects and linking clues all over the place, and I make all your choices communicable literally through actions you perform with items in the game (like MYST or No Man's Sky perhaps). But let's say it's a modern world. It's basically a modern sandbox; even if I wanted to, there's no way I could relate all the possible kinds of choices available to you, because they are as broad as the real world. This model is Object-Oriented. On a narrative control level, the best approaches will resemble a data-matrix of interacting values based on your choices and actions at each node, along with the current states of various other values both "internal" (like stats) and "external" (like objects), and the state of these data at any given time (aggregated, transformed, totaled or subqueried) is what determines the contents of the next scene. If there even are "scenes."
Those two games differ very little in expectations, but a lot (and very importantly from a design standpoint) in infrastructure. Still, for both games, since the common definition of "railroading" typically comes down to "plot insurance," an important design question is to what degree "railroading" as an artistic technique can be subsumed into the system and gracefully employed. And there are solutions, though they will differ very much between these two games.
Here's the thing: given that you can have similar expectations across very different games and yet enjoy both experiences, it follows that questions of Narrative Control are simply Design Questions. They are not shiboleths. Depending on what levels you wish to look at it, Narrative Control can be granted, withheld, divided, shared, constrained, automated, randomized, modulated, bought, sold, accreted, depleted, evaluated, accepted, ignored, or impossible. It may be determined differently in different situations or conditions, or by different values and different types of measurements (some quite abstract), or into different levels that run throughout the course of play (meta, macro, diegetic, micro), or into segments of different lengths, scope, and degree of narrative affect, or into different sets of related tasks, or many more ways that I haven't thought of.
In addition, stories are fractal in nature, and the good ones are multi-tiered. In other words, there are a whole lot of ways we can deconstruct things, and all of them can be written and mechanized in such a way as to affect each other. And the mechanics for all this stuff can be taken up in the design of the system itself, or it can be addressed in an ad-hoc fashion by the performance of individual GMs. Or both. So that's a design question too.
I assure you we are not overthinking this. We're barely beginning to think about this.