Gordon Watson wrote:(i) Proper matching can occur anywhere, even automatically, and certainly independent of both detector-operators.
Mate!? Think! Independent!?
Independent of the data collected by both operators? Please tell me how you can match the data without using the data? The matching is
certainly NOT independent of both operators.
(ii) "Logical nonlocality"? I won't be using that term any time soon. But (and not speaking of you) I know some confused people who do.
Perhaps you should start using it. Independent?! You swallowed the Bell hook. Dependence is not causation. Classical probability is nonlocal. It's an espsemic theory, not a physical one. P(A|B) does not mean B causes A, it means you need information from Bob in order to restrict the sample space over which to calculate the probability. P(A|B) is therefore, not a local quantity, it is global one -- aka nonlocal. Just because Alice(Bob) measured their outcomes locally does not mean P(A|B) is a local quantity. Matching is the same. You need information from both sides, in order to restrict the sample space.
(iii) I need no argument to say that "no theory is nonlocal"!
Sorry, you do. There are physical theories, there are epistemic theories and there are local and nonlocal variants of each. The problem, is that many have failed to appreciate the difference, and when they hear nonlocal, they think physical causation. QM is nonlocal in the same way as classical probability. But QM is not a physical theory. It's an epistemic theory just like PT.
You may be justified in claiming the world is not physically nonlocal, but you won't be forgiven for suggesting that there are no nonlocal theories. Bohmian Mechanics is proudly nonlocal. Path integral is proudly nonlocal. The problem I have is when people interpret epistemic nonlocality, or logical nonlocality as physical. You are making their jobs easier by confounding the two.
Perhaps that's why you thought I was advocating FTL earlier. FTL is strictly physical.
(iv) Re this from you: "The point being simply that it's wrong to talk of separated measurements ..." . Wrong? For me, I'd say: the better the separation the better.
With the lecture above, it should be clear now that spending millions to ensure separate measurements, only to destroy the data by nonlocal "proper matching" manipulation is utter stupidity. If the correlation was present in the in non-manipulated data, why should there be a need to match?