Now, of course player contribution matters.
Then it follows that sometimes, a problem will be solved primarily because of the player's attributes, and not their characters. And this is true in all avenues of play, not just social situations.
A PC might be envisioned as a mighty warrior, veteran of many campaigns, and tactical genius. The mechanics can
partially reflect that, but the player still needs to make sound tactical decisions during a combat scene, and if they don't, there's an unavoidable disconnect. So it goes.
Should a DM step in and prevent a STR 18 fighter from being a stupid and ineffectual combatant (the rules certainly don't). If so, for the life of me, I don't see why they should step in and prevent a CHA 18 character from being an ineffective face or negotiator. Or the reverse...
Combat resolution will have many, many choice points allowing that a decently intelligent character might outperform a more intelligent character. Skill resolution, OTOH, is binary. It's pass fail. There really is only one choice point.
Why are you assuming social encounters are unavoidably binary? That's not a good reading of any of the rules I'm familiar with.
A properly run social encounter should have as many decision points as combat, if not more, reflecting the give and take of natural conversation.
I think you can make an argument that there's traditionally been a dearth of good, thorough examples of this in the rules, but I'm not aware of any DM's advice saying social skill resolution should be resolved as single roll pass/fail.
What is being said is that the genius character should do things that the very intelligent character can't.
And they can. But the mechanical definitions say
nothing about what the character will
choose to do at a critical moment. The choice resides with the player. And if player choice matters, then you have to accept the genius PC might do something dumb, and the sub-genius PC might do something extraordinarily effective.
The untrained Cha 10 character can talk until he's blue in the face, but, he's going to have a MUCH more difficult time influencing people's reactions than the Cha 18 fully trained diplomatic character. The player's words shouldn't play into this.
The CHA 10 PC
should have a harder time, but when you go on to say the player's words shouldn't factor into the outcome, you are explicitly stating the player's contribution don't matter.
Or am I missing something obvious?
A better example, Mallus, would be a 10 Int Wizard and an 18 Int Wizard. Now, should the 10 Int wizard be regularly more effective than the 18 Int Wizard?
Are talking a 3e/Pathfinder INT 10 Wizard, who's limited to 0-level spells, or a AD&D Magic-User who can learn up to 5th level spells

? Let's assume AD&D -- it's what I'm running now.
Yes. It's fine for an INT 10 Magic-User to regularly be more effective than an INT 18 one. Depending on how each player decides to use their PC's spells. The game doesn't, well it shouldn't, play itself.
While I'm at it, let me ask you this, re: playing the character that's on the sheet...
How do you play a PC whose mental faculties far exceed your own correctly? I've asked this before and never gotten a good answer. To my mind, the only solution is to willingly accept the disconnect between the player's mental/social abilities and their characters.
Your average gamer will
not have the wisdom of Christ or the charm of Casanova, or in the case of some lucky rolls or a generous point-buy,
both the wisdom of Christ and the charm of Casanova. It's a given the player's words/choices will not accurately represent the character described by those stats. Again, so it goes.
I suppose you could get around this problem by reducing the importance of player contributions (and leave success entirely up to the character and mechanics)... but the game-y part of the game would suffer for it. I'm all for considering D&D as a kind of collaborative fiction writing exercise coupled with performance... but it's also a game of problem-solving (well, for a sizable number of gamers).
To resort to that last refuge of scoundrels, analogy: it's not much fun watching the New York Times crossword puzzle solve itself.