But that has nothing to do with 5e, or particular RPGs in general.
It's a feature (bug) of running published adventures.
For that matter, unwritten norms that exist in D&D (such as "don't split the party apart") aren't rules for a reason- because not all table play that way. Some tables are perfectly happy splitting the party apart! That was, in fact, somewhat common in the 70s and 80s.
The reason it has developed into more of a norm is just because of the desire to keep most of the table engage most of the time; it has nothing to do with meta-gaming.
For that matter, the rise of the norm against PvP likely mirrors the rise in people choosing to play their characters longer. When D&D had more meat-grindery features, it was less of a taboo because people were less attached to a single character that they might be playing for six months ... or six years. On the other hand, games that are single-shots, or rarely played as "campaigns" (or are explicitly about the PvP, like Paranoia) do not have the same norms regarding PvP. All that said, there are still table that don't use that norm.
Putting all of this in terms of "metagaming," if you're actually looking at "framing," seem ... weird. IMO.
Putting aside any value judgement on metagaming (I personally don’t care at all about it except for when it’s hostile to functional play…and the overwhelming type/amount of it isn’t hostile to play), I don’t understand your metagaming/framing dichotomy here. I’m assuming you’re talking about GM scene-framing or scenario-framing? There is absolutely going to be significant (possibly total) bleed in GM metagaming/framing such that their entanglement is not reconcilable; you’re metagaming to achieve the desired framing.
And I don’t understand how what is being depicted isn’t metagaming (again, putting aside any value judgment on any given episode).
I’m going to throw out a few examples here:
1) Player 1 and 2 (of 4) wants to split the party. Both of their characters have dramatic needs that involve going to
this place rather than
that. 3 and 4 are concerned that the party-splitting will negatively impact the collective payload (it will) and therefore put 3 and 4 at risk for the Adventure x they’re interested in. The GM is entirely neutral on the subject.
Ultimately, 1 and 2 eschew their dramatic needs and go along with 3 and 4 on Adventure x because concerns of a “hostile work (game) environment” + they’d spent too much table time disagreeing already. They’re subverting character interests/actions for concerns about the gamestate or the state of the play at the table because of the collision of divergent interests.
Whether play within Adventure x becomes a bit hostile and passive-aggressive isn’t relevant to the question of “was that decision-point navigated via metagaming?” Now if it did become hostile, that we be metagaming upon metagaming!
And you could make a case that both of them could be binned into “dysfunctional” (in order for play to survive contact with the enemy - divergent interests - one side must succumb to social pressure and table time issues in order to play at all). But functional or dysfunctional, it’s still players subverting their in character interests/premise in order to either play at all or to avoid a dysfunctional play experience (whether that is hostility or merely a crappy spread of 25 % play and 75 % no-play haggling).
2) The players have a plan to totally foil the GM’s master villain plans and send play careening into the PCs favor or end the conflict before it started.
Enter Scry > Teleport > Fireball by the villain.
Now it could be a thing that the GM has actually prepped this NPC with all the (a) non-systematized (let’s say it’s AD&D 2e so we don’t have stuff for this but we bolt on the kind of robust villain portfolio that is in a My Life With Master game in terms of dramatic needs and moves and means) stuff that players could uncover and act upon. Now let’s say the GM rolled for (got lucky!) the (b) NPCs available spells/items and (c) prepped the NPC’s standard spell and (d) henchman loadout.
Now that looks like a scenario that will be informed by very limited metagaming by the GM. The players can make moves to uncover the encoded nature of the villain (will they escalate to nukes…why wouldn’t they…what for), and all the intersecting means at their disposal (and how to Rock/Paper/Scissors it should it come to that).
Now imagine either none of (a) - (d) or a thing or only very partially (perhaps the GM has prepped a few henchman, a mild and mostly opaque motivator for mustachio twirling, and given the NPC level 17 Wizard).
One of these scenarios looks much closer to the GM pulling out the old Scry > Teleport > Fireball + Henchmen tantamount to flipping the table over in protest as they’re “losing” (control).
Noe they will surely fall back upon “BUT WORLD BUILDING AND FANTASY VILLAIN REALISM” in an attempt to legitimize and obfuscate the Calvinball play. But it’s the same thing.
Yes it’s “framing” (as in we’re in this scene now and play has taken this dramatic - anticlimactic really - turn), but it’s underwritten by metagaming. The GM isn’t acting upon encoded NPC stuff and playing with integrity. They’re going outside of that process entirely and vetoing a player gambit with an active block.
Why aren’t both of these underwritten by metagaming?
Again, I’ve long decried the GM move above (further, Ive got a problem with the nexus of free play + “but worldbuilding” fall back that even keeps that on the table..often players have no way of knowing the difference between the two…a combination of poor GMing, Table time stress, and system issues) while simultaneously not caring a wit about the overwhelming majority of METAGAMING FOUL laments.
