Additional comment to add: One other line of questioning comes to mind in the "what plot could we imagine where suppressing sociopathic / psychopathic / narcissistic traits in humans to the extent they are kept from amassing power leads to severe problems for society down the line" - what are people with such traits uniquely good at, if anything, which results in a net gain for society?

There's plenty of pop psychology which points out that people with at least some of these traits tend to be concentrated in roles like leadership (which may not be entirely due to self-selection) and first responders.

Just to look at one example, an additional trait (beyond limited prosocial emotions and willingness to violate the rights of others) which may or may not be necessary, and is not sufficient on its own, to constitute psychopathy is relative fearlessness. People with this trait typically aren't so fearless that they act completely irrationally with regard to self-preservation, but they very measurably experience far fewer of the effects of fear in dangerous situations. You can see how this may be useful among police, firefighters, soldiers, etc. How does this trait relate to the other traits in question? Does suppressing the other traits like limited prosocial emotions also suppress this trait, thereby depriving a society of General Pattons? And so on.

Hmm, that question of inevitable filling of the niche in a species with language is a good one, and one I hadn't thought of.

In part it seems to be downstream of questions about what sociopathy and/or psychopathy actually is, which have been in debate for decades and still very much are. "Limited prosocial emotions" is necessary but not sufficient. Willingness to lie and violate social norms established to preserve the rights of others at a cost to the actor, probably also necessary, I'm not sure (no coffee yet) if sufficient in combination with limited prosocial emotions.

My housemates have 3 cats, one of whom I would describe - as a cat lover who has lived with many cats the majority of my life and has one in my lap right now - as the cat equivalent of a violent psychopath. (I actually like that psycho cat and she likes me as much as she's capable of liking anything.) Since she's a cat, she's not capable of the sociopathic pattern of abusing and/or manipulating social structures at a high level, e.g. by deception, to further her aims.

She is basically limited to violence, either direct (you touch her, and if she's not in a good mood she shreds your hand) or instrumental as instrumental as an animal without complex planning ability can be (she stalks the cat currently in my lap who she hates, and maintains a more or less continuous campaign of suppressing fire to minimize that cat's territory in the house; if I leave the house for a few days, she quickly steps it up because I am the human in the best position to intervene in her suppression efforts).

In humans there is likely some distinction between those who are violent psychopaths with little to no manipulative behavior, versus ones with psychopathic traits (limited prosocial emotions, problems recognizing and thinking about certain emotions in others) who are manipulative and may or may not be violent. The question this brings me to is, is language necessary for such socially manipulative traits like the much-ballyhooed "superficial charm," and others like tendency to deceive, to arise? I'm not sure. And then subsequently, in a language-using species, is deceptiveness based on language and theory of mind - lying - for personal gain such an attractive niche that it must always arise?

I'm sure there's a lot lacking about my ramble here, it's early in my day. And I might be off topic. But I've been researching these traits as a layperson for decades because my life has been strongly impacted by people with psychopathic and narcissistic traits in several contexts, and thus I've developed an interest in understanding how their minds work.