I think the problem with the "inconsistency" as viewed by some is the fact that close states are not permanent, yet they are being viewed as a permanent state. This is simply not true; any post can be re-opened at any time. In light of this point of view, it should be clear that there isn't any inconsistency.
Note that close states also do not at all suggest that the question itself is "not physics." What it does suggest is that the question does not fit the site in its current form under current policies. A closed question may be a perfectly valid physics question; case in point: homework questions. We view these as off-topic, even though they certainly are valid physics questions. And since policies can change, there is no reason to suspect that an off-topic question will remain closed.
I do note though that, on occasion, there are questions asked here that aren't actually physics (e.g., some people asking engineering, philosophy or chemistry questions). This does not mean that the close state itself indicates that the question is not physics.
This also has been discussed on Mother Meta before:
The solution is mixed between the answers (so obviously no consensus, not even highly upvoted answers), but two points were brought up there that we ought to consider:
- What happens if we later decide that the closed-original is on-topic?
- What happens if we later decide that an unclosed-original is off-topic? What should we do to the questions closed as duplicate of that one?
In the first case, your issue disappears because the close-state isn't permanent. This should suggest that closing as a duplicate of a closed question should not present any problems/inconsistencies because of the fact that the original can be reopened.
In the second case, we have the opposite case you present. Should we then go through all the duplicates and re-label them as off-topic? That would be quite a lot of unnecessary work (because it brings us right back to (1)).
I disagree with tgp2114's evaluation process; the reasoning should be in this order:
- Has the question been asked before?
- Is the question off-topic (HW, engineering, etc)? Too broad? Opinion based?
- Can't tell or disagree
It should not matter if the proposed duplicate is also closed as off-topic because, again, it can get out of that state.