Pistonrager said:
FLAVOR IS MUTABLE! DON'T USE FLUFF AS CRUNCH IN COMBAT!
... (repeat ad nauseam)
Do you get it yet? Though flavor is important flavor is mutable... and honestly... have you ever had a DM destroy all the equipment something is wearing after the PC's dump fireball after fireball on it? Just because they fluff says something doesn't mean you can assume anything about it's combat use. Also... lead melts very easily... not a point of instant death for anything...
Fluff/flavor is not mutable.
Or at least, it's not mutable within a given campaign. If I say that in my campaign, Fey Step means you step into the Feywild-analogue of the mortal world, move through it, then step back, then that's how it works, and players have a right to expect that I will make rulings consistent with that. If a PC wants to know what's going on in the Feywild and announces he's using Fey Step to pop in and have a look, I can't just fall back on "This is just a short-range teleport ability, the rules don't say you can use it for cross-planar reconnaissance."
Or if you want a combat example--say the PCs are laying a trap for an NPC eladrin, and put a ranger in the Feywild with a readied action to shoot the NPC when he tries to bamf away. If I'm using the Feywild-movement fluff for Fey Step, then that plan should work, even though nothing in the crunch text says so. And it's entirely reasonable of the players to expect that plan to work, and to get mad if I arbitrarily announce that it doesn't.
And if
fireball says it melts lead, and I as DM accede to that statement, then the players have a by-God right to expect that they can melt lead with their
fireballs, in combat or out. And they can also expect to ignite anything else that would be ignited by a fire that hot. This may lead to some disagreements and require DM adjudication once in a while. Such is the price of playing an RPG and not a computer game.
Now, you can if you like deny the
existence of specific fluff for a particular mechanic--in other words, you announce that that mechanic is pure metagame and that you will make up fluff for it on a case-by-case basis when necessary. This is the approach commonly taken with experience points, since it's pretty dang hard to come up with fluff that makes any sort of sense for the way XP works. But this becomes silly past a certain point; if you're going to make up the whole story on the fly anyway, why are you bothering with rules in the first place? There
has to be a point of connection between the rules and the imagined reality of the game world.
Coming back to the OP's question... to me, the point of "rules aren't physics" is that the rules express how things
generally work; they are not absolute and exact specifications of the game-world reality, but approximations. When the rules produce results that make no sense in terms of the game world, it is the DM's responsibility to adjust the results as necessary (and perhaps apply a house rule if the situation is one that will come up often).
Now, this does not excuse rules from conforming
in general to the game-world reality. If the result of a rule has to be adjusted almost every time the rule comes into play, then it's a bad rule.
To draw a comparison with actual physics, consider the Newtonian and Aristotelian models, as applied to space flight simulations. Newtonian physics are a pretty good ruleset for space flight sims. Even though they don't actually match exactly what's going on, and there are corner cases where you have to adjust the results (e.g., when dealing with objects moving at close to the speed of light), in the vast majority of cases, Newtonian physics yields results consistent with what you'd expect.
On the other hand, Aristotelian physics are a horrendously bad ruleset for space flight sims. The results produced by Aristotelian physics are so far from reality that you will have to adjust them pretty much every time you try to use them, to the point that you might as well not even bother.
I would accept "rules aren't physics" as an excuse for using the Newtonian model. I would not accept it as an excuse for using the Aristotelian one.