I agree with this view. I think there cannot be one single governance body. At least not in the form that is proposed, as it does not fit the grassroots nature of how the fediverse came into being, and the diversity and culture of the fedizens that participate in it.
Saying “not in this form”, because there are evolutions of the Fediverse that can address the issues you mention. I believe that, when we are shaping online social interactions, we should model them to desirable social interactions that exists and work well in the real world.
You indeed already mention some of these, like Sortition. But this will not be the model that is attractive to everyone. It is essential to have Freedom in any way, shape or form to be baked into the social networks. Currently we have a very limited model, where you choose an instance based on its Code of Conduct (and maybe the appeal of the domain name), and then you have Followers / Following under your control.
Elsewhere on this forum in #fediversity:fediverse-futures I introduced the Community has no Boundary paradigm. It is an abstraction layer that allows modeling Community more as it exists in the real-world. Here people participate in many communities in various roles and representations.
Applied to the Fediverse it would abstract individual instances away, as well as the application domain the instance provides. It is more task-oriented. With this model an instance may come with a full Community identity, or be just infrastructure where you have a fedi account. Note that this is a good thing:
The terminology of “Instance” is a highly technical term, that has no real-life equivalent
It unnecessarily seggregates the Fediverse into instance-bound boxes. In the Community model you would fulfill arbitrary amount of Roles (and possibly online identities, with more extensions in place) in arbitrary amount of Communities.
Now going onwards towards “Governance”. Of course there’s need for communities to be managed, and certain rules (if only the CoC) to be enforced. With the requirement of full Freedom, each community should be able to decide their own governance. Subsequently - based on that - another community may decide what role members of the first community may play in their governance.
So in that way an ‘overall governance’ of the Fediverse might be spun up. It of course carries a lot of complexity, and will take time to evolve to something practical (there are frameworks to understand and model the complexity, such as Engaged Theory).
Now, in order to express these Governance models, there may be extensions on top of Community model, that provide e.g. Governance Policies, expressed in a Linked Data vocabulary. You attach them to a community, but the policies themselves are reusable e.g. a Sortition policy.
Note further that multiple projects are creating governed community software, such as Loomio. But each of these is doing things in their own way, contained in their own software packages. A federation model for Governance Policies would break these artificial application boundaries, and once again put Fediverse on a more task-oriented approach to extending its capabilities (where application boundaries disappear, are irrelevant).
Finally, I just created a new post Federated Moderation: Towards Delegated Moderation? where one extended option is Moderation-as-a-Service. In Community has no Boundary + Governance Models, this would just be certain people in certain Roles (e.g. “Instance Moderator”) that can be discovered via the federation model.