I can use the mechanics to prove you wrong. We all witnessed turns and attack rolls (mechanics) for the ants biting, associated with the Bite action on the creature stat block. We saw that the rolls had a given likelihood to hit, and where they hit we saw that they did piercing damage (mechanics). We all saw that the accumulation of piercing damage mechanically associated with the bites of the ants caused M's death.
But, HP don't equal anything. The only thing you can say is that the character lost enough HP to die. Piercing is a purely mechanical element that has no actual simulationist meaning. Losing X HP to an attack doesn't actually MEAN anything.
The example of the ants is being skewed by the fact that the PC
died.
Suppose that everything plays out just the same except that, when the PC has 3 hp left, some mix of good luck for the players and/or a GM-delivered intervention (the god of aardvarks turns up?) Another hit from an ant would have left the PC dying, but that hit never occurs. What has happened in the fiction?
clearstream appears to assert that the PC has been bitten nearly to death by giant ants. But there are many features of the system that belie that: the PC can get up and walk around as if unhurt; none of the PC's clothes or armour have been damaged; the PC does not have cuts or abrasions or scars from the ants' mandibles (despite the "piercing" damage type); the PC may be one low-level spell away from full hit points (if a low-level PC with 10 hp) or many such spells away (if a high-level PC with 100 hp); etc.
The same point arises if the PC falls to zero hp from the ant attacks, but then is stabilised by a friend using one of 10 uses from their 3lb healer's kit: whatever exactly it is that a relatively untrained person does with 135 grams of healing supplies in 6 seconds in order to stabilise the "dying", it's clear that this can't be the treatment of a genuinely mortal wound. In time taken and supplies consumed, it's not much more than the application of a band-aid.
These features of the system - that the narration of the consequences of being "hit" is hostage to what happens downstream with healing/treatment - are not accidental, either: they're crucial to making the death-and-dying, and recovery, rules make sense (just as they were in 4e) - it's a fortune-in-the-middle resolution system. (The oddity in 5e compared to 4e is the bit about the spells required to heal to full.) The system, at its core, is the same as the AD&D one, and Gygax relied on these very features as part of his explanation of how hit points work, and how saving throws (including saves against poison) work: those essays are probably the first published defence of fortune-in-the-middle mechanics, although of course Gygax didn't use that terminology. A more theoretically sophisticated, but at its core very similar, bit of advice on how to narrate consequences using this sort of resolution framework is found in Robin Laws original HeroWars books: because a bit like hp in D&D combat, action points in the original HeroWars ebb back and forth, and only when the whole thing is resolved can we tell whether that spear thrust that rendered Frodo hors-de-combat actually skewered him, or simply winded him temporarily.
If we can't know what has happened to the PC
now - in being hit by the ant's bite - until we know what has happened
next - does the PC survive, or die? - then the system is obviously not simulationist in the sense being discussed in this thread: its mechanics don't model ingame causal processes, and don't tell us what is happening in the fiction. They simply establish constraints around, and perhaps allocate permissions in regard to, the introduction of new fiction.
If HP don't mean anything, how could any game satisfy your definition of determining cause of death?
Trivially. By having mechanical processes that do tell us something about the fiction. As has already mean mentioned upthread, RM and RQ are paradigm examples.
**********************************
There seems to be some strange vibe in this thread, that the measure of
quality of an RPG is that it satisfy the simulationist design paradigm. That's why in this post I've pointed out that some important designers - Gary Gygax, Robin Laws - have deliberately used resolution systems that are not simulationist, and have written explanations of how and why they are not simulationist.
There's also another strange vibe, that all of those players - such as
@Thomas Shey and me - who stopped playing D&D back in the day because of (among other things) simulaationist preferences, were
wrong about D&D: as if it really does provide the same sort of information about the fiction, via the same sorts of resolution processes, as (say) RM. Which is bizarre: systems like RM, RQ and the like were
deliberate reactions to the lack of simulationist resolution in D&D combat (and other elements of the system too, like PC build). And Gygax's essays on hit points and saving throws, in his DMG, were deliberate defences of his non-simulationist design
against those reactions.
************************************
Consider an actual example. Looking at my copy of RM Classic Arms Law, I see that a giant warrior ant attacks with a 35 Medium Pincher attack. Let's suppose that two such ants attack a character who (via whatever previous turn of events) has been disarmed (and so can't parry), and who is wearing a chain shirt (but with no greaves or shield), and who has a +5 defensive bonus from Quickness.
The GM rolls for the ants: 36 and 69 respectively. Adding each ant's offensive bonus, and subtracting the PC's defensive bonus, we get totals of 66 and 99. Consulting the Beak/Pincher attack table, looking at the column for Armour Type 13 (chain shirt), we see results of 2 and 10AS. The "2" means that the ant has caused some bruising and surface level abrasion: the player subtracts 2 points from the character's concussion hits total, which is a figure that determines whether or not bruising and blood loss are debilitating you (upthread I posted some of its parameters: roughly, the range for humans is 20-ish to 150-ish in the conscious range, depending on the degree of body development, and 50-ish to 100-ish in the unconscious range - this is all "meat", to use the terminology of hp debates).
The "10" means that the other ant has caused slightly more serious bruising and abrasion, and the AS means that there is also a roll on Column A of the Slash Critical table. Suppose that roll comes up 58: the table says "Minor thigh wound. Foe takes 2 hits/rnd and must parry next rnd. +3 hits." So we know that the bruising/abrasion from the second ant's bite sufficient to cause 13 concussion points loss - enough to somewhat debilitate a feeble scholar but nothing that would set back Conan - and has also caused bleeding from the thigh, at 2 concussion hits per round. The rules provide further details on the debuffs to actions that flow from bleeding, how easy bleeding is to heal (1 to 5 points per round can be staunched using first aid, whereas 6 to 10 points per round are more serious and require more skilled non-magical treatment, or more powerful magic), the risks of a staunched wound reopening should strenuous activity be undertaken, etc.
Whether this PC lives, falls unconscious, or dies, we know what has happened in this 10 seconds of being bitten by giant ants. If the PC falls unconscious due to blood loss, we know where they were bleeding from, and also which ant delivered the more severe bite. The mechanical records of damage suffered - 15 concussion hits lost overall, plus 2/rd from the thigh - corresponds clearly to something that the PC is experiencing in the fiction - being worn down by bruising and abrasion, plus having a cut opened up on the thigh.
There are things that aren't specified - left or right thigh?, for instance. My experience is that this is sometimes dictated by circumstance - eg did the prior fiction establish which side of the PC the ant in question was biting? - and sometimes by random roll. Sometimes the crit tables specify (eg a 67-70 E Slash crit reads "Slash tendons and crush the bone in foe's shield shoulder").
Still, I think the contrast with "the ant hits; take 6 damage" is pretty clear. Likewise the implications: no "stabilising" the "dying" just by ticking of uses on a healer's kit; no in-combat healing that just as easily restores a dying person to consciousness as heals a light bruise on a child; etc.
Some people will look at the RM process and think it's awesome. Some will think it looks terrible - it's clearly slower, in play, than the D&D process. The point of this post isn't to advocate for it. It's to explain that it's obviously
different from the D&D process, and that that difference includes the mechanics providing - without the need for ad hoc or post hoc narration - an account of what is happening (not just "what has happened") in the fiction.