With that in mind and with other comments here talking about how these attacks mirror others, and were somewhat naive even in their methodology, maybe they intended to lose these accounts to just evaluate some Meta detection systems. Apparently Stanford is writing up a full report on this incident which I’m sure will be well studied by the military departments involved.

Depends on the impact distribution, I'd imagine? I.e they could go after the most impactful bad actors which, assuming a power law (typical for social networks), would have vastly more reach than other accounts. Or it could be that they're catching the low hanging fruit (careless accounts). Wish they'd share some numbers (i.e followers/reach of these accounts etc).