I think causality is often messed up if we are thinking just about what we do at the table. Resolving hit locations in RQ being a clear example. Within world, causality is fine.
That said, it seems worth making a list of proposed features identifying simulative mechanics, I've now
association between the mechanical process and the process that takes place in the fiction
provides information about how the result was achieved
results are not defined by a player statement and a successful roll
abstractions of a larger, more complex fictional reality
ease complicated processes into something that can be more easily played with
non-playful
change the fictional world to reflect character actions
Those last four are from a quick parsing of Sam Sorensen's "New Simulationism".
I mean, if you want a simple example: the complexities of narrating a critical hit in 3e (and to a lesser extent 5e).
You narrate that you are attacking the enemy. You roll a critical
threat--but, unless this is specifically 20, you must still confirm that you hit, as with any attack roll. (Which means, yes, this is a "simulation" which generates critical threats that miss in some cases. Don't ask me what that means physically, I have no idea.) Once you have confirmed that you hit, you must confirm that you really did get the crit. This requires you to roll and hit a second time. If you do, then it is a crit. If you do not, it is just a regular hit. Then, after all that, you roll certain damage dice (primarily, only those from the weapon itself), rolling twice as many as usual, unless it's a weapon that has a better crit multiplier (e.g. x3 instead of x2), then you follow whatever the multiplier is.
However...because you don't multiply all the dice...it is quite possible to roll very poorly on a crit--indeed, you can easily do
below-average damage on a "crit". By the time you're rolling the damage, however, someone has almost certainly begun narrating the attack action, meaning it's very plausible to get an epic description and then a wet fart of a result.
5e doesn't have the whole confirm-crits thing (and also has far fewer means of expanding one's crit range, and all crit multipliers are always x2), but is otherwise functionally the same. And I would know about that whole "rolled damage and got a wet fart" thing. Literally just happened the day before yesterday in Hussar's OotA game. I got a crit on
sorcerous burst, which is particularly nice because it has (limited) exploding dice...and then proceeded to roll a grand total of
three damage on 2d8.
It's one of the places where I'm truly baffled why they chose to go with the 3e way instead of the 4e way (and thus one of the reasons I believe 5e has dramatically more in common, in its design and its ethos, with 3e than with any other edition of D&D). 4e's crits are guaranteed to be a big hit--you maximize damage dice. Even if that maximizing-dice thing were only applied to the basal weapon damage, it would still mean that the highly unnatural "struck a telling blow for far less than even an average strike" thing didn't happen. A blow being a
critical hit would, in fact, be a hit that is critical--something anybody would be upset to have to deal with. And it isn't even like it's that
different! The average result of doubling XdN is simply X(N+1). It's extra work, for a
less verisimilitudinous result, that
often (as in, like, about a third to a quarter of the time?) results in worse results than if you'd just gotten a pretty average hit!