I can think of a number of ways:
I have less fun at the table if the players have less fun--even without the threat of players leaving. The dead silence of a table where no one is having fun is a nightmare.
The easiest way to keep the players engaged is to make their suspension of disbelief as easy as possible, to make it easy for them to conspire with me to keep the story believable. So, prior events absolutely serve as a constraint on me as the GM, both for narrative consistency and for ... judicial consistency, I guess (ruling the same way). While I don't think of GMing as being a perfect parallel to writing fiction, it's not unlike a writer of fiction wanting to maintain consistency and continuity over the course of a story.
The players who have given me information about their characters, have given me ways to tie their characters to the setting and to the campaign. If someone tells me, "My character's lover disappeared, and I want to find them," they've told me something about a kind of story they want to have emerge during play; if I want to keep that player engaged, I'd do well to have that at least as a repeating thread--and it seems like a good idea to have that resolve during play (in the instance I'm thinking of, the player has said he doesn't think the character would keep adventuring after finding her lover, so it either needs to be at the end of the campaign, or we need to work on a replacement character).
The players have signed up for a game where their characters are the protagonists. That is a constraint on the scenes and story-threads I frame in. The players have an expectation of fair play. That is absolutely a constraint on me. I think the fact some GMs act as though those aren't constraints is unfortunate--the horror stories are real, and they shouldn't happen ever.
As you might guess, I feel quite constrained as a GM, even in a system like D&D 5E that really doesn't mechanically restrict my options much.
This all looks good to me.
Very good post and I agree with it. You're talking about play moving toward the optimum possible for 5e (harken back to my 4 continuums of Protagonism upthread).
However, I think you're post directly above this more answers
@darkbard 's question than this one because I think (and he can correct me if I'm wrong), his question was about constraint in action resolution mediation specifically (not protagonism broadly).
So your answer to his question would be "synchronicity at the social contract level" (which is what I was intimating prior).
Its a fine answer (its "the" answer really).
I would say where the problem lies is the fact that "synchronicity at the social contract level" is an alchemy that is not reproducible on any mass level (otherwise we wouldn't have had the extraordinarily discordant output in the thread I mentioned and we wouldn't constantly be besieged by anecdote after anecdote of similar discordance on TTRPG boards...and in real life). Because alchemy like this isn't reproducible and is inherently volatile (you can have it and "poof" its gone because of a series of action resolution issues), it cannot serve to mollify someone who is looking for an answer from first principles that is stable and reproducible.
So, for instance, if someone simply told me:
"We're going to give you a panel of 10 GMs who are going to observe and peer review your 5e, level 15 game action resolution mediation for 10 sessions, and give you a grade A - F."
The LAST thing I would say is (and you're talking about someone who is extremely confident in their ability to adjudicate games) "yeah, we can just go ahead and forgo this whole thing...its a 100 %."
I would probably say "if I get north of a B-, I'll be VERY surprised."
And to me...B- is a failing grade. Because what that means is that I've had an enormous number of peer reviewers disagree with incidence of action resolution which would then aggregate into a hefty number.
So, the reality is, the alchemy of social contract at any given table relies upon several factors that go way beyond the competence and cognitive synchronicity of the participants at the table (eg - deference to perceived authority, respect for you, manners, no-effs really to give, conflict-averse personality, a hill one isn't willing to die on).
So I look at 2 realities conjoined here and what the implications one must draw from it:
1) There is a 0 % chance that any 5e GM in my theoretical peer review scenario above is going to achieve anything approaching 100 %...again, I'm way more competent than most and I put my median for any given set of reviewers at around a B-.
2) Social contract alchemy is x degree (with x not being an insignificant value) reliant upon "extra-competency and extra-synchronicity" factors and sussing out even the qualitative (forget quantitative) signature for any given table would require a major research project and neuroimaging equipment (and likely significant uncertainty in the findings still).
The only implications from the marriage of those two above and your persistent action resolution harmony at your table is "prabe is a good 5e GM and his alchemy with his players works."
That is a statement that all GMs should hope for as a broad statement of their play. But using it as a proxy for "a 5e GM self-constraining via extra-game principles smuggled in so that they can limbo well under their mandate + action resolution procedures can reliably achieve table-synchronous action resolution across any four 5e players (who are not inherently dysfunctional or combative)" is extremely fraught.
But honestly, that is intentful design. That is a feature, not a bug. The designers willfully designed in heterogeneity across the population of all 5e tables; "Rulings not rules, natural language, make the game your own, and find your own alchemy." But it doesn't stand up that you can reverse engineer that design intent to say that your "found alchemy" is reproducible at scale because "competent GM + intentfully designed cross-table heterogeneity." I'm not saying you're saying that, but if that is the implication, it can't stand up. I'm sure there are stray anecdotes of relative "Edens of 5e Protagonistic Play" sprinkled about the "5e-osphere" (like yours). But if its happening at scale, (a) its happening quietly and (b) all of the noise that says it isn't is somehow just a flukily robust noise masquerading as signal. Further (and again), where the anecdotes do exist, there is a lot of "extra-synchronicity" stuff that are consequential aspects of that alchemy (when it comes to action resolution mediation specifically).
Man, that is a lot of crap I just wrote. I hope that makes sense.