There is a limit on how big the gap can be, though.
The only limit is our imaginations, which makes the field theoretically infinite (but practically bounded, both by actual imaginations and by what we are prepared to cope with).
For instance, if you are going to have ingame causality modelled by the mechanics, then the mechanics have to model processes like the transmission of energy from A to B. This is what Edwards is getting at in his analysis of combat mechanics. And once you have those sorts of processes in your game, a minimum degree of realism is going to assert itself. (Contrast, say, Toon, where you can't have systems that model ingame causation in any literal sense because cartoon worlds don't actually obey causal laws at all.)
I would say it is entirely possible to play Toon using the system as a literal "cartoon physics" - again, it's all down to the scope of imagination. Even a world where "energy" manifests very differently would be conceivable, but difficult to make consistent and complete.
That is not to say that most players will not much prefer to play in a game-world with which they feel at least marginally familiar; Toon has help, here, in that cartoons are generally somewhat familiar. But that doesn't seem to me to change the fundamental truth that there are two quite separate mechanisms at work:
1) The desire to have the game-world fully reflected in the rules, and hence conversely have the physics of the game world defined by the rules (the rules may "dictate" the world, but the players dictate the rules used, so it really is two way - the determining direction is just a matter of timing).
2) The desire to roleplay using a setting (game-world) that is to some extent familiar and felt to be understood, which means including a healthy dollop of either "reality" or genre (or, most usually, both).
It might be interesting, too, to consider the role that the game "fluff" has in setting the expectations for the "realism" of the ruleset; it seems to me that setting genre expectations might be a vital function of "fluff".
I've never heard of anyone who thinks that peasant railguns are actually possible in the D&D gameworld, and relatively few who think that it is a stop-motion world. Yet these are both consequences of treating the 3E or 4e action economy as modelling the ingame causal processes.
In the sense that peasants could launch stuff into space? No, they really aren't, because that's not what the result of the rules would be; there is no "momentum" in these systems. Add some flexibility to the game-time "axis" (so that a turn length is related somewhat to the number of "chained" actions involved, at least in a field surrounding those making the chained actions) and it's not even all that difficult to envision.
More generally, I think you may well be right that many players don't really consider the implications of a "stop motion" game physics. When they do, a first instinct can be to call for "rulings", but that only actually introduces a need for another game system to be used, in effect - and that system will be very likely to introduce its own issues. So experienced players settle on ignoring the issue and handwaving it when it becomes intrusive, thus moving away from Purist-for-System.
Nevertheless, I don't think that either stop-motion or hit points make it impossible to have consistent world physics, and neither do I think that playing in what one accepts as a world defined by these systems but ignoring the fact as far as possible is actually disfunctional. Doing so and demanding certain other system elements "so that things make sense", however, might well be inconsistent.
My view is that, in fact, most players assume that motion in the gameworld is not stop-motion but more-or-less continuous, as it is in the real world (at least at human-sized scales). And that peasant railguns are not possible. On these (in my view, widely shared) assumptions the D&D action economy does violate ingame causality. Mutatis mutandis for hit points and healing.
For many players I think it probably does, or rather ought to in theory (if they didn't engage in handwaving/rationalising). It certainly introduces inconsistencies into their preferences and advocacy for other subsystems. But that is due to an (often stated) preference for "realism" (sometimes termed "believability", "verisimilitude", "reasonableness" or even "something that makes sense"), not due to a demand or preference for rules that model in-game physics.
Think of it this way, maybe: the rules of Toon, as written, model a form of cartoon reality. Few toon players object to them when playing Toon, but introduce the same rules to D&D and I'll bet many folks will be up in arms. The problem is not that the Toon rules don't model a perfectly consistent process - it's that they don't model the specific process that folks have decided up-front they want D&D rules to model. That is quite a different issue. It might even get to the heart of the "4E is not D&D" comments; to one set of folks, "let's play D&D" means "let us roleplay characters in a world that works as defined by the D&D rules", but to another set of folks it means "let's roleplay characters in a world that we envision as D&D - and we'll find some rules that approximate that world and alter them as required to fit our (GM's) view of that world". The first is PFS as I originally understood it from The Forge, the second is a variant PFS that says that "the rules" are actually what the GM (or players, in Universalis*) makes up to fit the (pre-determined) world model.
*:
Universalis is actually a fascinating case, since it seems to be built to facilitate exactly this approach with no GM.