Toxicity in games, and feelings of isolation

Well, I could write an in-depth analysis of each point on its own. But I feel like that’d be repeating the same general message over and over again, so I’ll instead break it down into the things that makes sense and what seems to be conclusions not applicable to WoW due to other factors at play.
You’ve also misinterpreted some things, judging by what you’ve posted in this thread. But it’s easy to do, when the report itself is signaling two different things.

To start with, the report is heavily focused on the preconception that our brains are stuck with a relatively fixed number of identifiable impressions we can handle.
But our minds aren’t that consistent. It’s actually more fluid, where it can remove past impressions to make room for new.
So with the dunbar’s layers, as I’ve mentioned earlier, while the numbers represent what we can handle in any given moment, what those numbers represent (as in which are actually good friends, casual friends and so on) can be changed.
Kind of like an inner popularity contest of where you’re drawn to, or end up encountering more often for example.

But here is where the principles and personality types comes into effect.
It’s not a secret that there are extremely passive people, or really aggressive people, or really amicable (as in easy to get along with) people and so on. This is the basic premise of personality typology.
What typology is about, is identifying as many of those typical traits and behavior quirks of the many types as possible, which that site I linked to have done a good job with. Even if you don’t take the test, as I’ve explained earlier, it just serves as a shortcut to a page where it explains the type.
You can reach the same page by going through the different types until you find one that speaks to you, if you don’t trust the website’s handling of the answers.

Another thing I haven’t mentioned before, is the IQ statistics. The average IQ of people in the world, is in my opinion actually pretty low (source of statistics: Mensa).
This is yet another factor that drives the creation of lasting primary groups. Secondary groups are less affected by these things, since they’re by definition impersonal in nature.
But basically, we seek that reciprocation loop (an example being a person you can have a good meaningful talk with, i.e. not superficial) so as long as the types aren’t at odds with each other (there are some that just don’t mix well in social settings), and the IQ differences aren’t too evident whilst talking, then it provides a good start.

IQ meaning the speed of understanding and how easily we can understand more and more complex structures and concepts, not what we’ve learned. Knowledge and education aren’t the same as IQ, although they can be indicators of an individual who performs at a higher level but not definitively. In other words, memorization is not the same as IQ, however a person with a higher IQ can typically learn faster (also depends on factors like what’s being taught). There are also other types of ability quotients, but I digress.
One example of a bad matchup of IQ and personality types, is a person who’s very ambitious and has a good sense for creating advanced puns on the spot, and so on. A person like that talking to a very passive personality type and who keeps taking a long time to understand each pun made by the ambitious fellow, would end up with friction. It’s like mixing two different types of energies and conditioned social norms.
Basically they’d not be very compatible. But this is an extreme example, and doesn’t represent the normal interaction of strangers. However, it does illustrate a point, that other “chaotic” factors plays a role too, which you can’t completely control nor predict in the social design of a game. You can only create systems around it, and ways to alleviate such occurrences.

As mentioned in the report, people also need a sense of familiarity and agreeing social norms to get a good flow of that reciprocation loop going for a long time.
The proximity principle is just that people are more likely to talk to each other the more they encounter each other after all, which is also referenced in the report.

The report also goes on to mention the formation of social hierarchies, how small intimate primary groups are fundamentally different from large impersonal secondary groups. But this is a step I don’t believe is needed to be addressed on purpose in game design. Without a proper understanding of the steps people take in the game to get to know each other, then you can’t just slap on designs everywhere meant to address the hierarchies and larger design aspects too. Because there’s a big chance if you go too authoritarian in the social design to tell people what the hierarchies should be, then it might feel counterintuitive to players. Basically the leadership given automatically in the automatic queues, the system just “tells you” who’s the one with the leadership role, without anyone actually recognizing it to be so, is one such example.

So personally, I believe that as long as a game doubles down on properly designing the initial steps of harmonious human interaction in the game, then the people can sort out their own social groups in the game instinctually as they go. People’s ability to adapt is usually quite strong.

I also don’t believe there’s a need to set the largest limit of community sizes to 1500 as the report claims (which would basically be a server community). Sure, our brain would end up swapping out one very low impression for another constantly. But the familiarity principle would still show its effect so people create their own perceived social spaces, even the impersonal side of it, the more they play. Think of it like a pub you frequent in a big city, eventually you and other regulars create your own social group. But don’t confuse this with guilds, because a guild exists with sub-groups and a shared goal, however with a “social group of regulars at the pub” then it refers more to naturally filling the need for friendships, which in game terms would be some people you naturally feel connected to that you often see at a congregation point in the game and often have common topics to share and chitchat about, just because that little group of regulars often have the shared interest that kept bringing them there in the first place. After that, it’s a process of “getting to know each other”, which as mentioned is influenced by types and social norms.
As mentioned earlier, an American would have a hard time connecting to people in this manner in Japan, because their perceptions of personal space is culturally very differently defined, and people in Japan highly value respect and culturally defined social hierarchies in any context. Which is an example of yet another possible factor relevant to the creation of friendships, which can’t be designed for when the game so willy-nilly crosses borders like this. Language barriers are another example that can disrupt the reciprocation loop as well as the immersive process.

But with that said, it’s not like I’m saying CRZ, sharding or the group finder as a whole is in any way, shape or form, good. It falls more under the definition of “throwing bodies together” as the report puts it.
There’s no sense of familiarity if you play with strangers from another server in a game like this. You may end up encountering the same person once in a blue moon, and may even end up recognizing the player, but your mind has already been conditioned by the overexposure of other interchangeable people who doesn’t say anything and you don’t say anything to them, that it pales in comparison.
There are however occurrences of social groups forming to reach a shared goal even with the game as it is now, but those are rare occurrences. One example being mythic+ puggers who ends up enjoying each other’s company and/or really like each other’s skills, so the goal of social satisfaction or to fill an ambitious goal is perceived to be better to do together. But this is not the normal, when bodies are so interchangeable like how it is now, which the report also mentions. This is also what I’ve noticed social butterflies excelling at, they’re good at drawing people in to play with them and to keep playing together despite all of this.

Right now, in the game, as far as the social design goes, the larger secondary groups are the most common social groups. Guilds sharing the common goal of progression, but with differences in ambition to complete it, so many ends up crumbling from the failure to maintain the social hierarchy and share of joy since many ends up being too impersonal.
The report also references this occurrence:

Many game designers assume that if there is a shared reward, people will naturally align their activities. This might work if humans were hyper-rational, profit-maximizing automatons, but they are not. Instead, players benefit from clearly-stated goals and examples of how they might work together.

However, the report doesn’t emphasize the “examples” as much as I believe it should, nor does it give proper examples of “clearly-stated goals”, because if that is done in an impersonal manner then that is yet again an asocial design.
A good example of this being the goals in a dungeon. It’s a clearly-stated goal, with the example of downing one boss clearly showing how to proceed with it, yet it doesn’t promote social bonding at all. I’d argue it’s when a group is presented with an “appropriate” level of challenge that it can help promote communication. But not with a timelimit, it’s actually better if it’s something that needs more time to defeat. Makes the entire experience more intrinsic and immersive. The game just needs to provide those principles and natural steps for people to consent to one another before starting a group activity, instead of leaving it up to automation because if people haven’t consented to one another then a challenge will only drive the group apart.

In PvP however, it’s literally a climate of strangers for most people. Even established rivalries dissipated over time, with the change of design direction.
Basically, with the ever-increasing reduction of skill differentiating designs and risk/reward structures, it turned it into a rather monotone gameplay experience.
And the one thing we’re exposed to the most as we play a game, is not the social experience, but the gameplay. So even with a good social design, without an appropriate gameplay design, then it’s still not likely to succeed.

The failure to uphold rules is also yet another problem, with the automated right-click penal system, it further emotionally distances a player from the gameplay, since it makes the glue of the “huge impersonal groups” which you can basically call the server’s community, where the glue is the perception of regulation, fall apart. The basic social norms don’t apply to such huge impersonal groups without reinforcing them, after all.

As for the design direction, from vanilla to the cata prepatch, there was the original talent template, with a few changes occurring here and there sometimes.
It was also in the ICC patch the dungeon finder (and subsequently automated teleportation straight into dungeons) came, and this patch brought a large jump in stats as well. I actually quit playing on my original account at this time, because I felt epics were given too readily in the new dungeons and the stats jump made it feel very overpowered.
With cata however, since the system worked in the way of providing talent points with levels, they felt people had too many points. So they reduced the amount of points in cata so it would only fill up a little more than one tree, while in wotlk there was enough points to fill much more than that.
So they compensated for that change by implementing a few new abilities and passive abilities, however they limited the abilities/spells to be specc-based at this point if memory serves me right. It was either then or with MoP, but I’ll presume it was with cata.
Then they replaced the entire talent template to the template we still have today, with the MoP prepatch. To compensate for this big change, they implemented many new abilities and passive abilities (passive meaning no need to press, they’re always active and some had proccs etc.).
But then in WoD, they felt they had gone too far and thus “the pruning” happened. While they did remove a lot and changed things, they also provided some more passives and an ability or two to compensate.

That’s what happened. But the REASON for why this happened, is that Blizzard first designed the encounters to fit the classes. So class designs took precedence in the design process, and encounters were made to adapt. But then they decided to reverse this flow of the design process, and started to design more impressive encounters but then had to make classes fit the encounters. This led to a simplification of class designs for the sake of empowering the encounters.

While I agree with that design style for PvE, that encounters are supposed to be the challenge, and less focus on the abilities of the classes, it’s the direct opposite of what PvP needs.
PvP needs room for skill differentiation, risk/reward structures, and a high climb for skill progression so the challenge never fades into feeling monotone.

Ghostcrawler felt that the biggest mistake WoW ever did was adding LFR, while another person from there has said that the biggest mistake was adding arenas, and so on. And looking at it from their perspective, then yeah, it makes sense.
But since PvP was such a huge success and enjoyed by so many for so many years, this downgrading of its huge focus on primary group creation once upon a time while now it’s a climate of strangers, and the challenging aspect being reduced so it feels monotonous for many, it is on a steep decline and has been for a very long time, when it has proven its worth already and should receive more attention.

So one of the big steps I think the game would benefit from, would be to separate PvE and PvP. Because the class designs are also designed with one affecting the other and vice versa, and so are the reward schemes with the PvP rewards taking the brunt of the compromises and so on. It’ll never be designed freely until they’re completely separated. PvE is also being slightly held back by the PvP side of the game, since you can’t go too far with the “matching classes to the encounters” philosophy while also making sure the class can at least be played in PvP. So both sides would benefit from a clean split.
They also mentioned in a Q&A back during Legion to explain the PvP talents and why the concept wasn’t taken further than it was, that it wouldn’t make sense to be playing the same character but feeling like you’re playing a different character just because of the content you’re doing. Which means PvP talents are a compromise where the PvP:ers will always take the backseat. Which, as mentioned, is another clear reason why they should just separate them entirely. To have PvP on specific servers without PvE, and PvE on the “main” servers. The feeling of reward progression and so on is still possible with account-wide achievements and rewards, after all. And it would enable proper designs for PvE as well as proper designs for PvP, simply depending on the server.

As for PvE, it’s kinda self-explanatory what needs to go. The game should be more narrative-driven in PvE, while properly respecting the familiarity and proximity principles. The dopamine-focused direction of PvE as it is now, with neverending grinds and 4(!) different difficulties of the same raids makes absolutely no sense whatsoever. To present grind challenges is one thing, to stimulate dopamine at the cost of quality is another.
One example of a grind challenge that was narrative-driven and world-building, was the attunement to raids and dungeons, like with the use of keys or magical attunements and so on. It didn’t suit everyone, but that can be blamed on other factors like not making it dynamic enough for such people. It did serve an important immersive role though, which is gone from the game.

As for travel distances, to reinforce the socially immersive process, they should’ve never kept old continents tbh. It would’ve been better if they would close off the old continent as the new one emerges with each new expansion, to keep both the content relevant, cost of resources down, the world more perceivable by the human brain (it can’t both fully explore the game over and over again while staying immersed with the game as it is now with the extreme amount of scenery, it falls victim to the same limit as with the Dunbar’s Layers with too many impressions so it ends up feeling bland overall), and would make it possible for cleaner social designs.
By keeping only kalimdor and eastern kingdoms, but having each expansion open up access to a new continent, you both have that familiarity principle of recognizing the “world” as being the original two continents, but the “something to explore” being the new continent of the expansion. This would’ve also made it possible to remove the chore of leveling constantly higher, and to just keep the levels at 60.

But automated queuing and psuedo-automated queuing needs to go for proper immersion.
The problem is that I think it’s too late. It’s better to start over with WoW2 and do it right, instead of trying to change the ingrained behavioral patterns players associate with WoW as it is now. Easier to condition people to new standards by rebranding it with a new lable and new surroundings, content and challenges.

Summary:
The game used to focus on intrinsic reward schemes and facilitating primary group creation with the clever use of the proximity and familiarity principles.
The game now focuses on extrinsic reward schemes and drowning the game with impersonal secondary group creations, seemingly under the impression that anyone and everyone can get along with each other, which isn’t exactly true. Relationships of different kinds requires different amounts of effort depending on the many factors and circumstances involved from person to person and the rate of compatibility, after all.

Oh, and guilds I’d argue are fine as they are, they just need to bring the server designs back to what’s socially healthy and remove all automated queuing, CRZ, group finder, sharding and so on. It’s essential for the sake of shared identity on a fundamental level, and for both the proximity and familiarity principles.
And of course, to split up PvP and PvE entirely to different servers with its own separated gameplay designs would benefit both the social harmony of being able to presume a fundamental shared purpose when on the same server, and give free reigns to design more immersive encounters in PvE and skill challenges for a constant sense of skill progression in PvP. The reward schemes would also be possible to improve a lot, since the reward scheme in one wouldn’t affect the other. One could argue it’d be two different games, but I’d argue it’d just open up whole new venues of explorative gameplay for both sides and effectively reinvigorate the game. Instanced PvP should also be limited to per server, the same as the PvE content, for the sake of those principles as well.

Queue times aren’t affected by the amount of players queuing as people think it is anyway. “More people = faster queues” is a common misconception, the same as rating pool sizes which can be arbitrarily manipulated by messing with the metrics of matchmaking and predetermined rating rewards/losses and so on. Yet people still tend to mistake “rating pool size = a direct reflection of player participation” which is also a common misconception.