Yea 5e is just as bad if not worse than 3.x with how much it rewards making a character with one or two good stats at no real cost for a dump stat or two. I think it's probably worse because of how bounded accuracy giving a flat curve and how the massive overuse of advantage combine to exacerbate it. My point is that the excessively generous 4d6keep3 allotment takes that a step further by ensuring that there is no available choice that might result in some other type of optimization to further endhriithr one true style.
Edit: the elite array is too good in the places a pc needs and perfectly adequate everywhere else so there is no meaningful loss, that's also an array that would be very over budget with a 3d6 based pointbuy
Yeah, I pretty much agree with you on all counts, other than the last one. 5e wanted things simplified, and it very much succeeded at that goal. Unfortunately, a fair number of the secondary consequences of that goal are not as desirable as the designers originally expected.
Personally, I think it is more interesting when there are enough points such that you can be good at whatever you're supposed to be good at,
and then also above-average at some other thing unrelated to what you're supposed to be good at if you so choose, or "so okay it's average" at all the other things. I felt 4e's system was pretty much perfect on that front. Even an "incorrect" race (note quotes) could easily achieve a 16 in its main attack stat, which might not be
great but is perfectly workable if you adapt to it. When you combine that the Fort/Ref/Will defenses taking the highest of two stats (Str/Con, Dex/Int, and Wis/Cha respectively), you get a situation where most characters will have at least one weakness (because you can't really advance more than 2 stats), but not so glaring a weakness that it becomes a massive liability.
13th Age also does some good stuff on that front via its Background rules, which allow for more flavorful, character-specific, and "natural" skill usage, rather than the sometimes clunky aspects of the existing system. E.g. it can be difficult to insert things like "Engineer" into the existing system, but "Former Chief Engineer Assigned to Drakkenhall +4" is perfect for that. Or the various and sundry things that "Imperial Drill Sergeant +3" or "Court of Stars Courtier (and Courtesan) +5" imply.
However, I don't think the problem is about what players are "expected" to have, it's that the rules have become deterministic (that is, you choose, instead of rolling dice) and highly focused on primary class ability scores, and so the sensible decision is to max out your primary ability score, which happens to be what nearly everybody else using the same rules is doing.
While I overall grant the point from the rest of your post (people settle on whatever they're allowed to take), two things:
1. In 4e, it was possible to start with a 20 in your primary stat, if you bought an 18 with your point-buy points and played a race that got +2 to that stat. In general, however, this was recommended
against for most characters, because you had to make too great a sacrifice. Few characters in 4e were sufficiently SAD to justify this, because the vast majority of characters wanted to advance (at least) two stats, usually an "attack and damage" stat (e.g. Charisma for Sorcerers) and a "utility" stat (e.g. Strength for Dragon and Cosmic Sorcerers, vs Dex for Chaos and Storm Sorcerers, and Con for Elementalist Sorcerers.) Hence, you generally went 18/16 as your top two stats, but it was perfectly feasible to do 16/16 instead and shore up your secondary stats if you wanted (that's generally what I did when I played Paladins, who are a bit messier in terms of what stats they value.) Hence, it is not ALWAYS the case that you want your core stat to be the best: it is possible to design a game where you have "two primary" stats, or one primary and one very important secondary, such that neglecting either is Unwise.™
2. I don't, personally, think that the shift away from randomness is a "problem," nor that it's particularly relevant for the end result. E.g., even in games where you roll 3d6 strict, in order, zero alterations, players will just choose to play whatever character suits those stats, e.g. if you got a decent Wisdom, you're almost certainly going to choose to play Cleric. If re-rolls or stat swaps or "roll and assign" are allowed, then all randomness does is randomly assign you a variable point-bought array, perhaps better, perhaps worse; you're still going to put your big stat in Charisma if you want to play Bard and your low stat in Int if you intend to play a beatstick who doesn't do all that NERD stuff like READING or LOGIC. Point being, the vast majority of groups had already gutted any of the meager benefits randomness might provide, and even when those benefits are enforced, player choice of character class accomplishes much the same thing as player choice of ability scores does today.
I don't think there's ever been a time when players didn't do this. It's simply human nature. Some folks will intentionally break pattern, but overall the pattern will remain. You aren't going to see a lot of Wis 8 Clerics, nor Fighters who have no positive physical stat modifiers. And I, personally, think that that's actually pretty realistic, unlike what some folks say. In the real world, you DON'T see a lot of people doing professional sports who are physically unsuited to the task. You don't see more than a very small number of, say, snipers with severe myopia. You don't see politicians who simply cannot do public speaking. People who suffer from dyspraxia or sever arthritis either don't try to play the violin, or have to stop doing so once their symptoms manifest. Etc. The realistic world is one where
most people who do a job are actually somewhat good at it, particularly if that job can be lethally dangerous if you fail. Exceptions will exist, of course, but they will
be exceptional.
The world puts selection pressure on "adventurers." Those who are weak at their core shtick don't last long. Over time, those who are good at it predominate. To do anything else would be unrealistic.