I disagree here; I think the article, its discussion, and its history *are* at the same level of importance.
I was thinking about it more in terms of dependencies between concepts. I see "History" as "history of the article", and "discussion" as "discussion about the article". So the fact that those links were inside the article was enough from my point of view to communicate the relationship (without the explicit "article" option).
From what you comment, I can see that this generates the need for a different way to present the way back to the article, which may introduce more complexity for "history" and "discussion" views and has its challenges too.
The problem is that some articles are going to literally have thousands of discussion topics.
I think that if it is worth it, we can deal with that. For example, we can reduce the scope to a time period (e.g., "100 discussions today").
The question is: is it worth it? I don't know the answer. I'm suggesting that if we identify that users are especially interested on some particular states of conversations/history (recent change, big changes, whether you participated...) we could try to anticipate those when they are relevant for our users.