There is no problem here because in practice MWI is nothing more than
the usual QM formalism to compute the outcome of experiments where you
then assume that the ensemble of all possible outcomes really exists.
Locality then follows from the fact hat the dynamics of the theory is
manifestly local. The Hamiltonian only includes local interactions and
observers are part of this dynamics. Although observer are not
explicitly treated as being part of the wavefunction that describes the
entire system, the assumption is that in principle, this is the case. In
practice, one can then proceed according to the usual QM formalism.
> That is all very well, but it is not a local account of violations of the Bell inequalities.
Well of course it isn't! Bell's Inequality has been experimentally shown to be violated, so if there are hidden variables they can't be local.
> Bell'e theorem applies in Everettian quantum mechanics in exactly the same way as it applies in one-world accounts.
Bell's Inequality applies to everything, even if Quantum Mechanics were someday proved to be wrong Bell's theorem would still be valid. Bell didn't need Quantum Mechanics to derive his inequality, he just needed logic and high school algebra. So even if someday something supplanted Quantum Mechanics that new theory would still have to conform to the fact that Bell's Inequality is violated.
>Bell's theorem proves that the effect is non-local, so no local account is possible in any interpretation of QM.
Exactly. And "local" means there is a limit on how fast information can be transmitted, and that limit is the speed of light. Many Worlds is completely consistent with that, in it there is no way you can send a signal faster than light.
> The other sectors are not just FAPP unoservable, they are not observable in principle.
Hugh Everett didn't wake up one morning and say to himself, gee it would be cool if there were a lot of different universes, I think I'll invent a theory that has an infinite number of them. Instead he asked himself what would happen if he took Schrodinger's Equation seriously and assumed it really meant what it said, and when he did that those other universes just popped up. The only way to get rid of them is to change Schrodinger's equation as GRW has done, or to do what Copenhagen has done and say that for some vaguely defined reason a vaguely defined thing called an "observer" doesn't need to obey Schrodinger's Equation. Many Worlds is just bare bones, no nonsense Quantum Mechanics with no silly bells and whistles pasted on. The only assumption Many Worlds makes is that the mathematics means what it says.