another huge point about why SYORTD doesn't really work in most D&D, not just 5e, and that's how the characters are actually challenged. D&D is built on an attrition model that's extended over multiple consecutive challenges (exception 4e, which was balanced to the encounter). If I want to put pressure on a D&D PC, I have to do this consistently over a period of time, and that time period is pretty uniform within each edition of play. This means that the GM must be able to control pacing in the game if this pressure is meant to come to bear, and if you're controlling pacing you're already outside of SYORTD. The games that do best with SYORTD are ones that allow for short-scale high pressure play that isn't contingent on pacing. BW, for instance, can go from zero to 60 very quickly, and PbtA is entirely structured on the snowballing effect of success at cost and failures being the majority of rolls. These systems, through their very structure, create pressure during play without having to do anything outside of play to introduce it. D&D lacks any real ability to do this.
This is a big thing for me.
To add to your remarks about other systems: what's key, in my view, is not that there are resources that get run down over time; this is a feature of Burning Wheel (eg sorcerous tax that reduces Forte; financial Resources that get taxed; supplies that run out) and Apocalypse World (eg each PC has a "wound clock" where damage accumulates). It's that pressure can be applied independently of the state of these resources. Eg in BW, casting a spell requires a check independent of tax; likewise acquiring gear: any action can be made to turn on a check, and it is the possibility of failure and the resulting consequences that generates pressure.
In a game where authorship is shared, mysteries are equally unknown to the group and as you say, why prefer the DM? Even so, the group must have something in mind that answers the question: "say 'yes', or roll?" It may be they cannot see that a declaration is valid, so as a group they might say "no" or as you put it require expansion to get to "yes" or "roll".
I find the phrase "shared authorship" a bit of misdescription for the sorts of RPGs that I'm referring to as using narratively/dramatically-based "say 'yes' or roll the dice".
The RPGs I'm thinking of, from my own play, are Burning Wheel, Prince Valiant (at least as I approach it), MHRP/Cortex+ Heroic, Cthulhu Dark (again, as I approach it), and 4e D&D. These all involve a pretty traditional allocation of participant functions - players who declare actions from the perspective and fictional position of their PCs, and a GM who manages the setting and framing stuff. The players don't "author" anything any differently from how an AD&D player "authors" that an Orc is dead by succeeding on an attack that reduces the Orc to zero hit points, or "authors" that the door is opened by succeeding on an Open Doors roll. or "authors" that the local shop has iron spikes for sale by telling the GM that their PC is buying some iron spikes and then, receiving no objection or a grunt or nod of affirmation, proceeds to adjust their money lists and their gear lists appropriately.
It is the GM who is empowered to "say 'yes'" - if the GM says "yes" when the player expected a check to be required then their is the risk of a bit of a dramatic fizzle or anti-climax, but the game won't break in any basic way. If the GM doesn't say "yes" then a check has to be framed. In these systems, the resolution of the check requires, among other things, knowing what the PC is doing/attempting in the fiction. If there is uncertainty about exactly what this is - ie, about exactly what the player believes that their PC is doing in the fiction - then that can be resolved via discussion: the Adventure Burner for BW has worked examples of this, and the 4e DMG has an example too, although less-elaborately set out. In both rulebooks, the focus is on being clear about what exactly the PC is doing, and how that feeds into the appropriate skill to be used to resolve the declared action (from memory, in the BWAB it involves stuff around Stealth and Acting and infiltrating an enemy camp; in the DMG it involves using Diplomacy in the context of traversing a desert).
This is quite different from the notion of adjudicating possibility of success, including auto-success or auto-failure, based on fictional positioning that has regard to known-only-to-the-GM elements of that positioning. I did
that sort of adjudication the other day when I ran some friends through a couple of hours of White Plume Mountain, using my variant AD&D rules. I think that sort of adjudication does not interface terribly well with a structured and systematic action resolution framework like a skill system; even in my AD&D system there were places where I had to make ad hoc judgements (mostly about when to call for "STR checks" based on a PC's open door chances) that were fine in a friendly romp, but that I wouldn't enjoy having to make in serious play!
As to whether D&D precludes the second mode. As written it does mechanically. An example being ability checks where players in most cases aren't empowered to call for one or set its terms. And as written it does narratively, casting DM firmly as storyteller. What is gained is much as
@Oofta describes. Characters live in the world, without knowing everything about that world.
In Prince Valiant and Cthulhu Dark, players can't "call for" an ability check. They describe their PCs' actions and the GM frames checks. BW is a bit different because its system gives the players reasons to
want to make checks, rather than just have the GM say "yes", and so the player is expected to push for them - though there is also both player- and GM-directed rules commentary about the need to avoid "test-mongering" that violates narrative-stakes-based "say 'yes' or roll the dice". In 4e D&D a player can call for a check in the context of a skill challenge - players need to succeed at checks in order to succeed at the challenge - but the check still has to be framed within the fictional context.
I make these points because I think it is helpful to be clear on where the difference lies between "say 'yes' or roll the dice", understood in dramatic/narrative terms, and the sort of 5e D&D approach being advocated by a number of posters in this thread. The deep difference is not in who calls for checks, nor in the relationship between fictional position and the framing of a check. It is in the relationship between narrative/dramatic stakes and the
resolution of the check.