Superposition of branches in MWI

[entropy1]

So I consider a measurement on a superposition, in MWI, leads to another superposition:
$(|A\rangle+|B\rangle)|Observer\rangle \rightarrow |A\rangle|Observer{A}\rangle+|B\rangle|Observer{B}\rangle$
If we come to the latter situation, a superposition of branches, why does that not mean that, since it is a superposition, and the probabilities add up to 1, that only one of the branches is real, since only one of them can have probability 1?

So I feel I am overlooking an elephant.

Edit: I guess the superposition is a consequence of the unitarity of the evolution of the wavefunction. But I think that such a superposition doesn't deliver reality as we perceive it. But that may be precisely the measurement problem.

 

[Demystifier]

So I consider a measurement on a superposition, in MWI, ... why does that not mean that, since ... the probabilities of ending up in a certain branch add up to 1, that only one of the branches is real?

Because, according to MWI, $\psi$ is something objectively real on an individual level, so a priori it has nothing to do with probability.

I guess the superposition is a consequence of the unitarity of the evolution of the wavefunction.

Of course.

But I think that such a superposition doesn't deliver reality as we perceive it.

I would agree with you, but MWI assumes that it does.

But that may be precisely the measurement problem.

Yes, that's one of formulations of the measurement problem.

 

[EPR]

So I consider a measurement on a superposition, in MWI, leads to another superposition:
$(|A\rangle+|B\rangle)|Observer\rangle \rightarrow |A\rangle|Observer{A}\rangle+|B\rangle|Observer{B}\rangle$
If we come to the latter situation, a superposition of branches, why does that not mean that, since it is a superposition, and the probabilities add up to 1, that only one of the branches is real, since only one of them can have probability 1?

If you assume that the probability of a certain probability outcome is 1, you will end up with just 1 possible universe.

But you don't have to assume so. There will be thousands of other vanishing non-zero probabilities. So you always end up with the MWI.