Let's have a short look at the Action Window flow charts on page 21 of the FAQ again.
We see that both step 4 (Passives) and step 5 (Responses) have arrows leading to a subset of steps numbered I-IV. This tells us that Passives and Responses also go through the same Initiation-Save/Cancel-Resolution cycle the normal action does, and that there's also a step where Passives induced by the Passive/Response in question are resolved. Such a step is obviously needed, because the regular Passives step has already passed by the time the Response is triggered. If it wasn't there, there'd be no point where e.g. CS Benjen Stark's passive ability could resolve if he was killed by a Response, say, No Quarter.
So far, so good.
Now, step 2 of the action window (Save/Cancel Responses) has no such arrow. This has to be due to space constraints, of course - we know that save/cancel responses can themselves be canceled, we do it all the time. Also, it is implied in the section on save/cancel responses on page 18 of the FAQ that saves and cancels also go through the Initiation-Save/Cancel-Resolution cycle, just like other responses.
So, if save/cancel responses go through the same steps I-III normal responses do (even if the flow chart doesn't actually have the arrow pointing to those steps), it stands to reason that they'd also go through step IV, right? That would mean that during step 2 of the action window, there's a separate step for resolving passives induced by save/cancel responses triggered during that step. That would in turn mean that there are actually passives that resolve before step 4 of the action window, no? For example, if CS Benjen Stark is killed to pay for Harrenhal, would his passive actually resolve in step 2.IV instead of step 4? That would make an actual difference in gameplay, so this exercise is not entirely academic.
Edited by Ratatoskr