Does the difference between contextuality, nonlocality and retrocausality depend on where we draw the boundaries? Suppose we have a quantum system and a measuring apparatus in a superposition of detector settings. Different detector settings would measure "complementary" properties of the quantum system. The act of measurement will occur at time $t_m$.
We need to draw a boundary between the system and the environment. There are boundaries in space, as well as a future boundary in time (think spacetime and histories).
If the apparatus lies outside the boundary, we have contextuality. If the apparatus lies inside the boundary, but the "final time" boundary occurs at $t_f < t_m$, we have nonlocality between the detector settings and the hidden state. If the apparatus is inside, and $t_f > t_m$, we have "delayed choice" retrocausality from the measurement to the earlier hidden state.
So, it depends contextually upon where we draw the arbitrary unphysical boundary.
Boundary = Heisenberg cut
Is this a valid "interpretation"?
 A: I didn't know until now what is the "Heisenberg cut". So, I cite here what I found in Wikipedia.
"In quantum mechanics, a Heisenberg cut is the hypothetical interface between quantum events and an observer's information, knowledge, or conscious awareness. Below the cut everything is governed by the wave function; above the cut a classical description is used. The Heisenberg cut is a theoretical construct; it is not known whether actual Heisenberg cuts exist, where they might be found, or how they could be detected experimentally."
I will use the above definition of "Heisenberg cut".
Now, oncept of contextuality, or contextual experiments is clearly explained in "Quantum Theory: Concepts and Methods", by A. Peres, Kluwer Academic Publishers, Chapter 7.
Let an operator A commute with two other operators, B and C, but B doesn't commute with C. One can measure A together with B, or together with C. If the result of the measurement of A depends on its context, namely on whether we measure A together with B, or together with C, the experiment is said to be contextual.
Thus, the concept of contextuality involves at least three observables.
Another thing: we can't place the Heisenberg cut wherever we please. It's not up to us to decide where and when the collapse occurred. Only when an actual measurement with a macroscopic apparatus is done, we can say that the effect of getting one result out of the different possibility, occurred. We can't say more than that.
Retrodiction: forbidden, leads to contradiction. See my example with Hardy's thought experiment, Does Hardy's paradox represent a proof against Bohm's interpretation of the quantum mechanics?
