I haven't read the whole thread, but Im going to jump in from the OP.
Agency in games is the product of inviolable rules which the players know and can rely on to achieve known goals.
This is correct, and this is referred to as volition. Volitional engagement is a wonderful thing, and something a lot of games, including RPGs, could stand to focus on building.
For example, most RPGs fail to:
These are also fair assessments, at least from examination of traditional RPGs. I've argued in the past that indie RPGs also have similiar issues, just through different vectors.
In general a lot of these problems stem from, as I argued, the non-acknowledgement that RPGs are fundamentally a hybridized improv game, and without recognizing and designing to account for it, you can break the improv loop, and this results in all the odd idiosyncratic problems we think are unique to RPGs, which are really all just different forms of blocking and other improv problems.
And, of course, poor design work on the non-improv elements, but often the two are interrelated.
One critical area of divergence between RPGs and most other games are that many RPGs don’t clearly state the goals of play.
This goes back to the issue of volitional engagement. Regardless of what the game states as its object of play, volition is fostered by players feeling empowered to identify and set their own goals within the game.
Much of the time, this kind of engagement can be intrinsic to the player, and thus the game just has to have captured the interest of the player for them to be volitionally engaged.
But, this is tricky to guarantee, and doesn't in turn give you any way to design the game to support whatever they come up with.
By designing for volition, you can identify what is and isn't going to be volitionally engaging about your game, and then you can design to support that volition. And, in the case of RPGs, if you design with improv in focus, and create a robust enough system, you can actually support a lot of that intrinsic volition.
But, at that point you start defying the idea of what RPGs are the more you open the game up to this. That was a conundrum I faced, and
I eventually concluded I wasn't building one anymore.
The fact that a roleplaying game is played in the imagination doesn’t change the requirements for agency. We can play chess without a board and pieces, purely in the imagination too. In the absence of inviolable rules which the players know and can rely on to achieve known goals imaginary chess falls apart exactly the same as over the board chess.
The imaginary component of RPGs does, however, provide additional cover for claims of agency in games which feature very little.
This is ultimately why I've argued that RPGs are fundamentally improv games, because what you're pointing out is the exactly the breaking point of the improv game. The moment one participant blocks another, the entire dynamic collapses and the game has to be disrupted to bring it back on track.
The thing about RPGs however, is that participants aren't just people. The game is a participant as well, and if we assume a GM is present, this creates a three way improv dynamic that's incredibly fragile if not designed for, and most often RPGs have favored two of the three to the exclusion of the third.
This is why these improv problems manifest in idiosyncratic ways across all kinds of RPGs, because when you hybridize two different kinds of games, you need them to actually integrate with each other.
RPGs don't do this with improv, or do it poorly, and even the ones that try don't do so with transparency, which in turn is what I think keeps the whole hobby niche and relatively inaccessible, even as people go down the minimalism rabbithole.