1. Very much this.
2. Fixed, specific time intervals can be overly restricting - turn-based combat and fixed-length combat rounds being the poster children for this - and can (and IMO frequently do) conflict with point 1 just above. Preferably, if an action takes x-amount of time to do or attempt, that's how long it should take in the fiction regardless of game-play concerns. (in D&D I've seen in these forums a few homebrew attempts at sort of rolling-initiative systems that eschew hard-coded combat rounds in favour of actions just taking as long as they take in a more granular measurement such as seconds or - in 0e-1e-2e terms - segments. From what I can tell these ideas add more complexity than the typical D&D player (or DM!) is willing to live with, but I can see the benefit behind it)
Definitely I agree this can be a problem with turn-based combat resolution, and that's an area I'm used to compromise. Immersion generally suffers in combat scenarios, but that's often a trade-off for ease of resolution and sufficiently interesting gameplay abstraction. Like all aspects of game design, immersion is in tension with other design goals and you'll have to make trade-offs at some point.
Outside of turn-based resolution though, this has more impact on things like skill rules, particularly things like stealth rules, where you must specify the circumstances that call for checks ahead of time. For an easy example, consider a check to open a locked door. I need to know as a player what the costs of trying that action are (time, potential lost resources), what the potential failure points are (how many checks could be called for, what are the potential results of a failed check) and what my general chances of success are, so that I can make the best decision about trying the action. A game with an unbounded fail-forward system (or a system with particularly bad "success at cost" outcomes) will fail to be immersive, because the potential downsides of a failed check may make my decision making as a player not align with my character's putative lockpicking ability.
3. So, no meta-currencies. Got it. Love it.
Also, no meta-narrative resource management. A 1/encounter ability must be tied to a resource intrinsic to the character that refreshed on say a 5 minute time scale, not to an "opportunity" that they see arise in the opposition, because the decision to use the ability must be something both the character and I as the player can choose to do.
I get this, but at the same time I posit that if fully immersed character motivation ideally takes precedence to the point that player motivation becomes irrelevant: the character does what the character would do even if doing so runs directly against the player's interests or desires. (somewhat extreme example: the player really wants to keep playing the PC but the PC is faced with a situation where due to its established history and personality the in-character thing to do would be to permanently sacrifice itself for the survival of others; ideally here the PC sacrifices itself and the player just has to live with playing a different character henceforth)
This is fair, but I tend to think of character motivation as defining the victory condition of the game you're playing at any point. My motivation and decision making as a player is then based on trying to optimally achieve that objective. Personal survival is generally a reliable character goal, but if it isn't anymore, I can still as a player try to make good decisions to achieve whatever goal has superseded it if the character's motivations change.
Agreed. But what if doing the anti-narrative thing is in fact what the character would do, even though you-as-player know it isn't what you want? My usual example here is a situation where a player role-plays a character right out of the party; not necessarily due to in-party conflict, but due to the in-fiction fact that being elsewhere and-or doing something else is simply what that character would do at the time.
An example for a good-hearted Ranger type: "It's about a three-week round trip to take these rescued people back to safety, they're in bad shape, we can't take them with us into further danger, and if we leave them here in the wild to fend for themselves they're hosed. We're on a clock and can't all spare the time so I'll take them out myself; meanwhile you guys carry on without me. I'll see you back in town."
And with that I've just role-played myself out of the party (and the game, unless I can bring in a replacement PC somehow) for at least the remainder of that adventure.
This feels like a situation you'd have to accept some abstraction, not at the level of mechanical resolution, but at the level of character motivation. You're adding another goal to the character's motivations that mostly goes unstated "remain part of this group" and you might be prioritizing it over other motivations that might otherwise make more "sense" for the character. It's back to the abstraction necessary to play a game at all, the compromise here being that you agree to play as part of a troupe, even if that might not otherwise align with your motivations.
Agreed. If it ain't a challenge, what's the point?
Absolutely, but I'd push that challenge in this sense is a necessary outcome of the scenario, not the system itself, if immersion is to be maintained. This goes back to what I was saying about discrete timeframes earlier. If an action I can declare as a player has an impact on the narrative outside of the direct result (say, a mechanic that allows me to succeed by creating a future complication token that the GM can invoke to add another threat later) I cannot immersively take that action, because my calculus as a player cannot align with the character's decision making.