Dear Sylvain,
Thank you for keeping your promise that you will "read and refute" my article, and to continue our "dialogue". Unfortunately, you are again missing the point and misinterpreting what I write :)
Let me explain what I am doing in the article you discuss, and in others: I am exploring the consequences of a modification of the standard interpretation of quantum mechanics, under the assumption that the unitary evolution is never broken by a discontinuous collapse. Some reasons why I am doing this: 1) there is no direct evidence of a non-unitarity in quantum mechanics, it was only inferred from the outcomes of the measurement, 2) a discontinuous collapse would break the conservation laws, and this also was never observed, 3) a discontinuous collapse introduces a tension between quantum mechanics and general relativity.
One of the consequences of maintaining unitary evolution even during the collapse is that the initial conditions of the measurement device and the observed system have to be correlated in advance (as I proven mathematically here). Hence, the initial conditions have to be very special, like in contextuality. This may seem for an observer embedded in time, assuming she would know the complete information about the wavefunctions and not only the outcomes of the measurements, as being retroactive. However, by looking at the complete solution from the "outside", as in the block world view, these correlations appear merely as global consistency conditions, pretty much of the kind of consistency condition which Schrodinger used to find the atomic spectra from his wave equation. I explained this viewpoint more in this talk and this essay.
Please note that I don't consider that I am the keeper of the absolute truth, I am merely exploring a theory, because I consider it a better alternative for several reasons from which I mentioned some above.
What I do in that paper is that I explain that, if unitary evolution remains unbroken, then a succession of two incompatible measurements still can be explained without discontinuous collapse, if both of the two measurement devices perturb the observed system in a special way. You misunderstood this, when you said "what your claim rigorously means is that the first measurement device (that made the first measurement) keeps physically interacting with the system during the time between measurements, so as to progressively align it with the basis of the second observable.". No, I don't claim that it keeps interacting with it, only that it perturbed the observed system in such a way I described there. You also said "You are confusing between (the mathematical expressions of) entanglement and physical interaction." You are incorrect, I know very well the difference between interaction and entanglement. What I say is that we can see this as a superposition of composite states (which is entanglement), and each composite state contains a different interaction. So your misunderstanding is clear again.
You quoted my article that "The configuration after the branching has to evolve, so that the agent can see where it is going", and replied "Where does that idea come from ? I see no reason for it.". Well, as I explained there (where I was talking about free will), to make a choice, you need to know what that choice means, some of its consequences. Otherwise what kind of a choice is that? The alternative is letting the randomness choose for you.
I want to point out that my unitary interpretation is compatible with a version of decoherence, and also with relative state interpretation. If each of the decohered branches is unitary and not plagued with the discontinuous collapse. But usually the density matrix is given a meaning at the beginning (as representing the state), and another one after is diagonalized (as representing a statistical ensemble representing the branches after decoherence), and this amounts for each branch to a discontinuous collapse. Regarding your comment to my statement "I find hard to believe that the environment is the cause of selecting the eigenstates", I suggest you to take a look at the delayed-choice experiment and see if your explanation still holds. Since you called me incompetent for not agreeing that decoherence solves the measurement problem, let me remind you that also Roger Penrose and Anthony Leggett criticized it. Anyway, although I find your personal attacks not conformal to the rules of the contest, I prefer that your comments will not be removed.
Cristi