I'm not terribly concerned with the attributes representing real human (or almost-human) capabilities. Yes, there can be people who are strong, but not tough, or vice versa. How many of those do you see in fiction and gaming? How many of those have enough of a discrepancy for it to matter? In a way that can't be represented with other elements, like talents or class features?
With all due respect, I'm failing to see your reasoning behind being unconcerned with with attributes representing human capabilities. What would you have them do, instead? Also, when you ask "How many of these do you see in fiction and gaming," is that supposed to be part of the reasoning? "We shouldn't allow for these characters on the basis that no one ever allows for these characters"? That's a bit of an infinite loop. The state of things justifies the reason to make the state of things how it currently is.
I don't like trying to do this purely with talents or class features. Unless you make a talent group that's like "Endurance 1, Endurance 2, Endurance 3," etc., you just wind up with 2 options: You're character is equally as tough as everyone else who's of average toughness, or your character is exceptionally tough (just like all other above-average toughness people). That's not very exciting in an otherwise robust world. And having a spectrum of each talent like that would just be redundant, when your attribute system was already allowing for that.
You can tune it all you like, but without a gradient, you just create binary options. The people with the different value, and the people with the base value.
Like I said above, merging constitution and strength makes sense because neither of them is good enough to stand on its own, if we're talking about traditional ability spreads. Edge of the Empire by Fantasy Flight Games uses a general Brawn attribute and it work well enough. Traditional strength is of very limited use to people who don't rely on brute strength. Constitution is everyone's second choice. Combine them and you've got an attribute that competes with more versatile ones.
I don't doubt that Brawn works "well enough." That doesn't change the fact that there are large functional differences between "Stronstitution" and two separate attributes. I'm also not trying to say that there's no merit in the argument that Strength and Constitution are not 100% entirely separate things. I'm simply trying to illustrate the utility of each of them, as systems have used them in the past AND as systems have yet to use them. I'm not saying "do it just like someone already did it, or don't do it at all." I'm all for new designs and mechanics.
Also, as I've said before, it depends on what the game wants to do with the metrics. If your game is never going to use just-plain Strength much, then sure... don't use it. But, if your game is going to use something like Brawn, then just still represent both things (such that everyone who's strong is super high-endurancey, and vice versa), but the game world doesn't actually want everyone who's strong to be identical in other regards, then the attribute is failing the system in a way. Measurably, and not just "I like this more", etc. There's nothing wrong with preferences... I'm not trying to belittle preferences, but they are secondary to objective design goals and the achievement of those goals.
And yes, Strength isn't super valuable to everyone. But it shouldn't be. If you want a weak character, you should be able to make a weak character. The whole "attribute viability" problem is when Stat A isn't viable for Class X. If I'm a Wizard, and I have no use whatsoever for 1 or 2 attributes, that becomes a problem. That's what they technically fixed with Pillars' system. The problem is that they made everything affect the exact same things for all classes. And/or that the classes are too restricted in what kinds of things they can and cannot do. That's a bit of a side-topic, though. But, Constitution is a go-to stat because staying alive is always good. In fact, usually, the only reason you WOULDN'T want to spend points in Constitution is when you choose a class that gets piddly HP bonuses from Constitution points (be it initial HP or per-level gains). So, in my mind, that problem is with the HP/class design, and not with Constitution as a basic character metric.
I hear ya, but honestly... you're either going to do all the math no matter what because you need all the uknowns to be knowns for your min-maxing urges, or you're just going to ride the intuitive flow. If you're more casual and have an interest in wearing some heavier armor for its style/armor value/what-have-you, then you're just going to go "Oh, okay... the heavier it is, the slower my actions are or the more stamina I use whilst wearing it and doing things (etc.)". That's a pretty intuitive relationship. "Do I want to be a bit slower but have heavier armor? Yes? Cool."
It's really not that big of a deal. I mean, if that's a problem, then attribute points are a problem from the get-go, as every point you put into ONE stat is a point that another stat lacks. I don't see that as a design problem, but just as a side effect that means you have to do a little bit of work the more picky you are about exact values. Any system should be intuitive enough for people to just pick what they want without doing a ton of math IF they're not already inclined to do all the math and be super particular about their values. This is one way in which the d20 DnD attribute values work well, though. If you add one more point to Strength, and you know Strength determines your chance to hit with melee weapons, and you know you roll 1-20 to try and hit things, then you can intuitively comprehend the value of +1 Strength in that regard. If plus one point of strength affects your chance to hit by 3%, it's not as clear. Now you have to math, unless you're a math-oriented genius-brained person, in which case, go you.
Having worn mail, I can tell you its not strength you require but endurance. Fatigue is the real killer, especially if its warm and you don't keep yourself hydrated. If you are not used to it, a penalty to doing things like walking on a tightrope too as your movements tend to be more exaggerated and you feel like you are walking under water, but that is something you can get over with experience.
I understand that, but if you abstract things out into numbers the way that game code has to, it wouldn't actually be PURE endurance. If you drop 40lbs of chainmail on a tiny person with no strength, they're going to collapse in like 10 seconds. Even if they can jog for miles under their own weight. Again... this doesn't mean that if you have a 300lb Strong Man Competition guy, that he can definitely wear full chain and perform physical activities for hours on end. So you're right in that just checking Strength for stuff isn't really cutting it. But, you can't really separate fatigue from strength, as it plays a part in that.
It's a bit like Perception or Dexterity/Agility completely deciding your precision on their own. You might can see really well, but not be able to aim where you're looking, or vice versa. Going back to Strength and Endurance, you see this all the time in Ninja Warrior. That's the best example I can give. Some contestants have the strength necessary to overcome many of the obstacles with ease, but wane around the middle of the course (or fail to hold themselves up through a whole obstacle). Others can go for days climbing around with their fingertips/hanging/swinging/etc., but hit certain obstacles and lack the strength to get past them (they can't lift themselves high enough, etc.). It's two pretty clear factors. Are they completely separate? No. Are they the same thing? Not at all. What's the best way to represent them in an attribute system? I don't know. That's an excellent question that we should discuss, I feel, but I don't think disregarding these truths about everything helps any attribute system in the least.
Many traditional systems don't do it perfectly, but I don't think discounting everything about the way they do it is the answer.