Benders are clearly powerful individuals who can do impressive things impossible for non-benders.
There are still extremely important non-bending characters throughout AtLA. Sokka, Ty Lee, Mai, Suki, Piandao, the Mechanist, etc. Characters who have special (but perfectly mundane in-setting) skills, training, resources. And in a world where spirits are very real and the supernatural is but a breath away, even mundane acts can have effects beyond what is possible IRL. And yes, there is a TTRPG for AtLA, so this isn't just a narrative to game comparison.
The problem with referencing the importance/contributions of non-powered individuals, insofar as it comes to them existing alongside powered individuals in narrative media, is that you'll continually run up against the same issue: the narrative has been deliberately constructed in such a way that it essentially goes out of its way to justify their presence. It's how you put Batman on equal footing with Superman (to use a perennial example where this sort of thing comes up).
That's a lot harder to do in games that aren't built to support narrativism from the ground up, i.e. there are a lot of meta-mechanics that players can use but which don't necessarily correspond to any sort of action taken on the part of their PC. Where verisimilitude is a primary concern, the differences between what powered characters and non-powered characters can do quickly becomes rather stark, which is why we
keep having so many
threads about how to
square that circle.
Now, you
can still have such a paradigm under a more "gamist" paradigm as well, because in those types of games the differences between powered and non-powered largely become moot. Fourth Edition was a great example of this, where the nature and differences between power sources made little practical difference in the course of play; a warlord could restore hit points by yelling at someone just as much as a cleric could by casting a healing spell. But that breaks fairly hard from verisimilitude, which is why a lot of people had a problem with that in the context of D&D (which has always had a large amount of verisimilitude in its presentation, though it's varied between editions, and has always had areas where it compromised in that aspect).
Hence, this whole thing becomes rather circular. Verisimilitude requires limiting certain things and being completely hands-off with others, but the decision about what to be hands-off with is either totally capricious and arbitrary, or knowingly and intentionally biased. The "veri-" to which there is "similitude" is a choice, not a requirement.
The salient principle to keep in mind is that verisimilitude's central principle is building an in-character understanding of how and why the game world works in the way that it does. Once you've established those principles, everything is extrapolated out from there (including exceptions to the rules, once you've defined why they're exceptions). In that regard, nothing is capricious or arbitrary, short of what rules you've set up to begin with.
In that regard, a verisimilitudinous model is "arbitrary" only in that those rules had to be established in the first place, but that's no more "arbitrary" than someone designing a gamist set of rules ("the pawn can only move one square ahead...except on its first move; then it can move two" is an arbitrary decision) or really any other set of rules which govern a game. Sam Kass made the arbitrary decision that rock-paper-scissors was too limited, so he came up with rock-paper-scissors-lizard-Spock, arbitrarily stopping at five choices instead of three. Why not seven? Or nine? Because arbitrariness!
Likewise, the fact that certain things are limited and others aren't available isn't anything particularly unique to a verisimilitude-based model either. All games have those, and all are enforced under a model called "the rules." Obviously, rules can be modified to the satisfaction of the people involved in a given instance of play, but that's not really here nor there. The point of this thread is to look at the virtues and advantages of a set of rules based on in-character design principles, as those allow players to intuit how the world works from their character's perspective, and so allow for a more immersive (and, by extension, more fun) experience of play.
Folks who like the things volunteered to get the short end of the stick aren't exactly keen on that. Being told, "oh, but you see, giving those things short shrift is actually necessary for the good of the game!" when we can clearly see that it is simply a choice, in support of a very specific and narrow set of preferences.
Again, this isn't really an issue with verisimilitude; all games necessarily make certain choices about what is and is not allowed. You can't have you queen skip over pieces in chess, and people who'd prefer to can see that it's simply a choice not to allow that, a choice in supprt of a very specific and narrow set of preferences. To say "I don't care for the set of ideas that this model is built around" isn't to levy any sort of charge against verisimilitude in particular. At best, it's simply saying "this type of game isn't for me," which is fine, but it's not pointing out any kind of weakness or drawback unique to that mode of engagement.
"Verisimilitude" is simply a more subtle, wily version of "realistic." As soon as you start pushing on the specifics, its true colors are revealed: only sanctioned breaks from reality are permitted, and all other things must conform, not simply to IRL physics (which are quite a bit more flexible than many realize...), but to the rather stunted subset thereof which is held in popular opinion ("pop physics"?) as physically real and consistent. Actual physical feats real Olympic gymnasts, archers, and swimmers achieve are usually impossible or effectively so under this stunted subset of our physical reality, to say nothing of mundane-but-epic feats fantastical characters should have within reach.
I'll refer you back to the OP, where I noted that "realism" in the context of TTRPGs is typically used as a shorthand for "how things function in the real world." Now, as I likewise said, the problem is that it's very often
also used as an attempt to discuss the easily-intuited-but-difficult-to-articulate idea of internal logic and/or self-consistency. In that regard, talking about "realism" is something that I think invites misunderstandings and lack of clarity in what's already an area fraught with a lack of common terms and definitions for things that a lot of us can clearly pick up on but have trouble talking about, especially when there are differing shades of understanding and nuance. To that end, there's nothing "subtle" or "wily" about verisimilitude, except that it attempts to more accurately reflect the difference between "realism as how the real world works" and "'realism' as how the game world works."
I very much appreciate that folks want to feel that a world is grounded, sensible, predictable. A place explored less through map and word, and more through action and consequence. That is why I included "Groundedness & Simulation" as one of the four game-(design-)purposes" in my possibly-incomplete taxonomy. But I find "verisimilitude" goes quite far past "I want a world that is grounded, sensible, and predictable." Coupled with various rather pointed preferences and the advocacy for certain design trends (which, notably, were not true of classic editions; loot in early editions specifically favored Fighter characters over other classes!), it amounts to a shell game of the problems of "realism." Advocates have recognized that "realism" is a problematic term, so many of them have cordoned off the problematic parts and just don't speak them out loud...until pressed.
One thing to make clear here is that "groundedness" (and related terms such as "sensible" and "predictable"), which is one of the primary benefits of a verisimilitudinous model, is not the antithesis of either "supernatural" or "transmundane." Quite the contrary! One of the things to keep in mind is that discovering how the game world operates is quite often one of the primary goals of engaging with games designed around that principle, and as a natural extension to that, trying to figure out how to reconcile seemingly-contradictory things (if and when they happen) is one of the major areas of engagement. We see things like this quite often in narrative media, often when the protagonist(s) encounter some new villain with special powers, and there's some exclamation of "(s)he can use two different forms of magic at once?!" or "they can cast spells without an incantation?!" or "they can resist that irresistible attack?!" etc. What typically follows is an investigation into the how/why of it, followed by either a countermeasure or the heroes realizing that their existing knowledge was wrong/incomplete and training up their powers accordingly. It's verisimilitude that gets them there.
You can have a world that is fully self-consistent, and where explicitly-flagged "magic" is not the only way to exceed the bounds of what should be physically possible. The supernatural and transmundane are vast. They contain multitudes.
Which makes you sound like an advocate of verisimilitude yourself.
