Moderation and centralization while typically aren't independent, aren't necessarily dependent. One can imagine viewing content with one set of moderation actions and another person viewing the same content with a different set of moderation actions.
We sort of have this in HN already with viewing flagged content. It's essentially using an empty set for mod actions.
I believe it's technically viable to syndicate of mod actions and possibly solves the mod.labor.prpbl, but whether it's a socially viable way to build a network is another question.
It seems far too risky to sign up on a service for the purpose of intercommunication that is able (or even likely) to burn bridges with another for any reason at any time. In the end people will just accumulate on 2 or 3 big providers and then you have pseudo-federation anyway.
However all the LGBT+ friendly servers federate with each other and that's good enough for me. I like not having to see toxicity, there's too much of it in the world already.
"People will just accumulate on 2 or 3 big providers" is far from an inevitable circumstance, but there are conditions that make it more likely. That, too, is largely down to negligence or malice (but less so than the abusive communications problem).
Is that still true? As the admin of a small instance, I find the abuse coming from mastodon.social has been really low for a few years. There is the occasional spammer, but they often deal with it as quickly as I do.