Any game that wants a ton of content has this problem. The many caves in Skyrim start to feel the same after a while. You can often recognize the same assets over and over in very content-rich games. The total number of assets used in the article is tiny compared to what a production game would have. Procedural generation can't be used as an excuse to do less work, or it just feels same-y. A game is still a ton of work; it's just different work for different goals with procedural generation.
Generating a huge world is boring if you don't have interesting objects to put in that world. Interesting objects are meaningless unless they're connected to a story. Minecraft works because it's a sandbox for you to invent your own stories, so the infinite world ends up getting filled by the player. If any element feels repetitive, the whole game will. If you wanted a fully procedural game, you need to figure out how to procedurally generate all the different aspects: story, characters, enemies, places, objects. It's more work to do that well than to make a traditional game. So people going into procedural generation thinking it'll save them work are already destined to fail.
I think Path of Exile does this kind of procedural generation of areas very well. Generated maps of a specific area will feel similar with some landmarks and the direction you're traveling in, but this is a good thing, because they can be used as guide posts where to go if you're an experienced player.
PoE2 on the other hand is some of the worst map generation iv'e ever seen. Like a copy and paste of the same exact thing hundreds of times to give you a large lifeless drudge of a map.
> Procedural generation can't be used as an excuse to do less work
> So people going into procedural generation thinking it'll save them work are already destined to fail
Consider the possibility that nobody really picks up proc-gen in the hopes they can laze out a RDR3 or such over the weekend.
Another thing is, this applies to indies and AAAs alike: while a big world has to have interesting unique things in it, by definition not every square meter can/should be chockful of another "interesting truly unique thing" because if the whole world is filled like that, it's just another kind of sameyness in that the novelty factor would wear out just as quickly once you're getting that there truly is true novelty in absolutely every little square meter, which kills the novelty sensation in a heartbeat. Novelty delights us in a backdrop of routineness, sameyness, same-old-same-old-ness. So in between interesting things, the thusly necessary slightly-"duller" in-between areas are to pace and prep and make one anticipate novelty "hopefully almost just around the corner". Ideally it's so spaced to appear just in time before the player resigns such hopes.
And so if you're going to have slightly-duller "filler areas" (and let me posit that any real-world say forest (in a biome one has traversed before), without the physical air and smells and winds (or friends/pets stringing along) is quickly "proc-gen dull-ish" within minutes — even in reality — or call it "meditative"/calming) — so again, if you're going to have slightly-duller "filler areas" just to connect and space apart the unique content things to good effect, then procedural placements/scatterings/variations are going to beat manual placement not just in "effort time" but because manual would swiftly look much more repetetive (being inevitably eventually effectively copy-paste driven) given the scale of environments under discussion, even if it would not take a human months of menial clickery.
Rockstar gets it right imho but their approach would also get your "many caves in Skyrim" right because they manifest little novel uniquenesses not in scattered objects or env textures/models but via lively interactive interludes, whether it's animals frolicking or chasing each other or attacking towards you or "random stranger" incidents etc. That's the right kind of "filler but not boring", nobody cares about the variety of the rocks, only noticing them if nothing seems to be happening.
Generating a huge world is boring if you don't have interesting objects to put in that world. Interesting objects are meaningless unless they're connected to a story. Minecraft works because it's a sandbox for you to invent your own stories, so the infinite world ends up getting filled by the player. If any element feels repetitive, the whole game will. If you wanted a fully procedural game, you need to figure out how to procedurally generate all the different aspects: story, characters, enemies, places, objects. It's more work to do that well than to make a traditional game. So people going into procedural generation thinking it'll save them work are already destined to fail.