The problem is not that some people want to imagine a world in which physics are way more complicated because of magic, the problem is that they want to fit magic into the currently understood laws of physics, and it simply is incompatible, just as the black body problem was totally incompatible with newtonian physics until quanta were thought of.
I want to fit magic into the currently understood laws of physics, or at the very least have it be able to seamlessly operate alongside said laws, and so I gave it some thought and came up with a few new laws of physics that apply in magical places of which - due in part to those same laws - real-world Earth is not one.
On that chassis I can build whatever I want.
=====
As for realism in general, two things:
First, the OP redefining the word "realism" to mean something other than reflecting real-world reality doesn't help the discussion one bit, as now we have to come up with a new term to fill in where "realism" just was: the idea that the game setting has a real-world (or at least real-universe) grounding and that, using the principle of specific overrides general, things work there as they do here unless something says otherwise. The advantage of this is that a whole bunch of very basic things can simply be assumed rather than written out, in that the rules only need to deal with - and be - the "something says otherwise" parts.
Someone upthread mentioned game-world gravity as being something that isn't realistic due to falling damage. I counter this by suggesting gravity in all ways other than falling damage works by default just like it does here, and that it's the rules-based abstraction of falling damage which is faulty.
Second, I think the underlying premise that says realism-simulationism-whateverism is simply an attempt to force reality into the game system is backward, in that it's assuming the game system takes priority and that when they come into conflict it's the in-setting reality that has to bend to suit it. I prefer to look at it the other way around: that the game system in its entirety is simply an attempt to model and abstract the in-setting reality and thus when there's a conflict it's the game system that has to bend to suit because the priority is to preserve the integrity of the in-setting reality.
Yes there will be unavoidable places where the in-setting reality cannot be reflected well by the abstracted game system and-or where the game system has to have its say in order to remain playable, and that's fine. But where there's a choice between system (or at-table convenience) and in-setting reality, I say go with the in-setting reality and in so doing make the whole thing more believable and consistent. A good and easy example is fireballs that conveniently pixellate themselves into 5' squares - sure this makes things easier at the table but as it makes no sense whatsoever in the setting I'd choose not to use this model.
Same thing with tracking rations, torches, ammo, etc. - yes it's convenient at the table to ignore these things but doing so violates setting integrity (and, IMO, good-faith play; but that's another issue entirely) and thus I expect at least some attempt at tracking them.
