It's less than an ethical split, actually. If A does not federate Threads, but B does, Threads still does not meaningfully impact the experience of users on A. No defederation between A and B is needed for A to maintain their desired experience.
As such, there isn't a split. There's an ethical difference, but the impact is negligible, and thus it doesn't require disassociation, which would be what an "ethical split" would be.
Until recently the fediverse took pride in the fact that they watched out for eachother. If tgere was an instance that didn’t moderate nazis, they defederated or at least muted it.
Or if they were Beehaw, and the other instance got too big. lemmy.ml soft-blocked HTTP requests from the KbinBot. And so on and so forth. Add in all the drama that went down in Mastodon between instances. You're painting a very rosy picture of a tidy, well-behaved Fediverse when in reality it's been pretty messy.
Not that this is relevant, as mentioned above.
Now, that the instance in question is run by a corporation with a history of bad moderation, desinforamation and hate-speech they get the benefit of doubt, because (...)
Again, this isn't relevant in the context of causing a split. Let's assume Threads is full of Nazis. 100% of users are Nazis. No! 200% of Threads users are Nazis!
None of those Nazis will be able to get content onto A in the earlier example, at least not from within Threads. If A wants to block Threads, they can just do that. Blocklists don't have to be common between other instances, it literally doesn't matter.
Thus [Meta] will not let the rest of the fediverse become competition.
Meta does not have a way to impact Fediverse projects without the consent of the project they attempt to impact. They cannot "stop" Mastodon or Lemmy or Kbin in any way. It's FOSS.