Sorry, but I don't think you're going to sway anyone with that argument.
Well, in fact, there is a whole edition of D&D that deploys the very idea I stated.
Appeals to special relativity shouldn't be necessary.
The point of the appeal to special relativity is simply to show that one property, which is constant, might be the function of two properties that are context-relative. In the case of special relativity, the space-time interval, which is constant, is a function of location and time, which are relative. In the case of a harpy in 4e, the toughness, which is constant, is a function of level and role, which are relative.
Thus, a high level minion, and a low-level solo, can have the same toughness, although the mechanical mode of expressing that toughness has changed: the high level minion has high attacks and defences and low hit points, the low-level solo has low attacks and defences and high hit points.
I mean, we are trying to create a simple model for the purposes of gameplay, right?
The 4e model is pretty simple. I've encountered many posts by people on these very boards describing how they use it. None of them seemed to find it hard to work with.
Not only is the 4e model not very complicated, it does not create any inconsistencies either: the harpy's toughness is held constant.
Regardless of what reality is causing the +AC or +HP or whatever, and regardless of any debates or opinions on the matter of what represents what, you still have the consistency that X undefined property of the harpy results in it having 31 hit points. You can't change the mechanical reflection of that reality without either altering that undefined property or introducing inconsistency.
This simply isn't true. I can change the distance between Dr Watson and Sherlock Holmes, at a certain time and place, by speeding up Moriarty. (But the Holmes-to-Watson space-time interval won't change.)
Likewise I can change a harpy's level and hit points by changing the level of her opponents. That doesn't change the constant reality of her toughness. It just changes those things which are the context-relative expressions of that toughness.
As I said, given that in the real world we have context relative properties of which constant properties are functions, yet the real world is (presumably) consistent, there can be no basis for denying the feasibility of rules that work along similar lines.
You may prefer to think of hit points as a constant, caused by certain hidden variables that you can't identify. But there is no basis on which you can argue that that is more consistent than my approach, or the only pathway to consistency. (And that is before we even get to the issue of what it means for a
purely fictional entity to contain hidden variables within it. What do you tell the players when their PCs start doing experiments to discover these "hidden variables"?)