I found that a few questions I have answered recently as "new" were in fact duplicates of old questions that were not answered at the time, or may accept alternative answers. How should I handle those "past duplicates"? Should I mark them as duplicates of the new questions, or should I mark the new questions as duplicates and reply both in the older one?
We typically close the one that doesn't have any answer (not necessarily the old one) as the duplicate of the one that has an answer. If the existing answers don't cover a particular detail, we also add another answer to the original one as an alternative.
If both of them don't have any answers, then the more generally asked one might be the one that should be left open for a broader audience.
But please keep in mind that sometimes it is not possible to decide because each covers a particular aspect so we can keep both open and answer them differently. There is really no point of closing if they are both answered and linked.
So closing is not necessarily a good thing since it blocks the possible more general answers in the future.
Linking it however, even in the comments, snaps it to the next of the question in the Linked area and that is sufficient for the trail tracking (normally if there is no cross-link, then that field is invisible).
Regarding, the linking, as I am more active in Stackoverflow lately, I'm basically noticing this problem much clearer that sometimes you can follow a 6-7 chain of duplicates randomly closed though none of them answer the originally closed question.
By the way, I really think that our users who think that we should adopt SO rules should go and hang out a bit there and argue again here later. You can't even find the same question after a minute after it's asked if you don't remember anything about the title. It is raining questions. That place is fundamentally different and I don't think that we can speak about a SE network run by the rules that SO people set up. That's just impossible. So adopting their practice is not in our best interest.