I think the secret sauce there is that they're roguelikes. They have meaningful combat and they have the potential for wild builds that completely trivialize that combat. So why does this work for them? Because you can't guarentee a specific broken build every run. They're short and random, so the likelihood that you will put together all the pieces needed for a specific build before the end of a run is fairly low. By contrast, while ARPGs are "random", they're played over such a long term that it's expected that you will be able to acquire exactly the things you need for your build eventually. (Outside of chase items, but those usually aren't build defining for that exact reason.) PLus a lot of your build is defined by entirely deterministic mechanics. You get to choose your skills and passives. And with trading you can take nearly all the uncertainty out of whether or not you'll be able to put together the remaining pieces.
So because it's expected that you'll for sure be able to build what you want given enough effort, if you optimize your build to trivialize the game, you'll always be able to do that. When you get a a broken build in a roguelike, it's because you high rolled that run and you get to have fun experiencing the high point relative to the baseline. You know how tough the combat usually is, so the fact that you can now breeze through it without thinking about it is fun in and of itself. But if it was always like that, it would just be a boring game. Incidentally, this is why I tend to not like roguelites that allow you to define a lot about your build before you enter the run itself. They make it a lot more likely that you break the experience in a very predictable way.
In ARPGs the high point is the baseline. Either the game is able to be trivialized with a good build, in which case it always will unless you go out of your way to nerf yourself, or you can never really make the game easier no matter how good your build is, in which case the build making isn't super relevant. There's a reason people joke about Fashion Souls. The gear you can equip is often so pointless that you might as well just pick armor for how it looks.
An interesting case study for a sort of in the middle experience that kind of illustrates some of this is Noita. For those unfamiliar it's a roguelike where you play as a mage/alchemist descending into the depths of the world in search of mysteries. Your builds consist of wands that you can put an assembly of spells and modifiers in to craft very different spell setups. You also get some perks occasionally that do the usual kinds of things you'd expect from a roguelike passive item system. The game is brutally difficult to a degree that's deliberately unfair to the player. Enemies are chaotic. The environment is volatile and filled with things that can kill you in an instant if you're not careful, or even if you are careful because some enemy triggers some flying thing on another screen that flies into you out of nowhere. Many spells in the game can hurt you too and even the ones that can't directly can sometimes have a firing pattern that will make it hard to avoid hitting explosives and stuff that will kill you. Healing is extremely limited. Early on the game is certainly very skill based in the sense that you aren't going to immediately break the game in the first level or so, so you need to be able to avoid things while you slowly kill them. If you really enjoy build crafting, the early game is fairly boring in that respect. But ultimately as you progress it's more knowledge based. Your will be hard pressed to outskill later enemies if you're still running a dinky no damage wand. So you kind of have to find ways to break the game if you want to succeed.
SPOILERS beyond this point:
That's the initial experience. Two things become true once you learn more about the game:
-
There are a handful of very powerful combos that are way better than most of what you can do in the game. Once you know about them, either through discovery or from reading about it online, you will kind of ruin the build potential of future runs. You can somewhat reliably find at least one of these most runs so long as you make it past a certain point. There's not nothing cool to discover after that, but they're all way less practical and only something you will be able to do once you've already reached a point where there's no challenge they're needed to overcome.
-
Upon freeing yourself from the initial core run to go see the rest of the world(s), you gain access to essentially unlimited perks. You can gain absurd amounts of health, damage reduction and healing, immunity to a lot of hazards, enough movespeed to avoid most things, and the ability to basically get anywhere you want on the map, etc. You basically become a god of death and destruction, untethered from mere mortal concerns.......... until you randomly get turned into a sheep and die instantly. So similar to a broken ARPG character, you reach a point in the game where the only things that the game can possibly do to threaten you is to strip you of everything that makes your build and just instantly kill you. And similar to an ARPG, this only really happens because you can play a run for many hours after the initial, more roguelike length run.
There's probably something to learn from all of that if you want to try to thread that needle, but I think it at least shows the challenges of reconciling the tension between mechanical skill and cool build making.
I suppose part of the conversation is about a concept I call "difficulty pressure." (Maybe there's another term for it?) Essentially, how does the game's difficulty affect players' approach to optimizing builds in a game with them.
When a game is on the really difficult end of things, (and this goes for competitive multiplayer as well where the "difficulty" is that all other players are optimizing and you need to be better than them to win) the game pressures you to optimize your playstyle in order to just survive and overcome otherwise insurmountable odds. In this extreme environment, there sub-optimal builds get pushed out even if they seem fun because you will very likely fail with them. Thus limiting build diversity.
At the lower end of difficulty, the game might be so trivial that ANYTHING works, but it won't feel satisfying because nothing you do really matters. You probably don't even need a real build at that point, so that feeling of making something crazy that trivializes otherwise challenging content isn't there. There's just no reference point to appreciate how good your build is. If every enemy had 1 HP, without damage numbers, how would you even know how much damage you were doing? A build that did one damage would be the same as a build that did a trillion damage.
Like you said, ideally there's some good balance state where things are challenging enough to serve as a yard stick, but there are still a lot of builds that can reach that point. There's a boring way to achieve this easily: No builds. Or at least no difference between builds. Everything does the same thing but maybe the colors are swapped around. Obviously that's not really what we want out of an ARPG, otherwise we'd just play a pure action game. So builds have to be different enough to allow for very different experiences, but not so much so that some are essentially invalid. But that's a much more complicated problem. With so many pieces and combinations, it's virtually impossible to balance faster than the internet hive mind can optimize.
There's another boring way to achieve this: Not on the player side, but on the encounter side. Because a very wide variety of playstyles need to be able to complete the content in a roughly equivalent way, the challenges need to be relatively interchangeable because you don't know exactly what tools the player will have access to. So you flatten the content so there aren't sharp edges that will make some builds unable to beat it. Alternatively you can require the players to have a specific set of tools no matter the build so that they can deal with all these scenarios. For example, in Noita, you pretty much always need:
You have 4 wand slots and you will usually need at least one empty wand slot to be able pick up new wands in a level unless you can meet some other specific conditions. So all the slots you can use to make your build are spoken for. This limits what you can build a lot. Late game you can combine some of these effects into a single wand, but until then you have that restriction.
If the game didn't have this variety of challenges, you'd be more free to choose what you want out of your build, but then the actual content would be way less interesting.
This is the core tension. Content asks things of you and your build is the answer to that. The more difficult or specific the challenge, the less freedom you have to make different builds. The more generic or easy the challenge, the less your build matters, meaning you have more freedom but it's less satisfying to act on that freedom.
EDIT: I forgot to discuss the action/skill axis: Some games, despite having builds and being mechanically difficulty, can be entirely overcome with skill. People do challenge runs of Dark Souls at SL1 with a broken straight sword and no armor because fundamentally, nothing in the game requires you to take damage or kill things at a certain speed. So you technically don't need a build. Skill is essentially all that matters. If you just avoid things forever it doesn't matter how little health you have or how long it takes you to kill something. Any build stuff you do on top of that is just for the fantasy and to make things go faster. (Although fantasy wise I think the souls games kind of do a terrible job. All the flashiest weapon skills and spells are way too slow and impractical to be useful. They're not just suboptimal, some of them will actively make the game harder than if you were using nothing at all.
I guess my point isn't that it's impossible to make a game that has elements of both. It's that they are inherently antagonistic, not synergistic concepts. The more builds matter, the less content does and visa versa.