I'd suggest a two phase approach. 1) come into conformance with the least possible change to current experience, and 2) make the structural change which might change current user experience but provides for assisted access improvement. I'm not a fan of big-change, and I am of a belief most of the current HN mob are pretty OK that it remains much as it is (not that they are unwilling to countenance change but there's a certain curmudgeonly love of this being an un-lovely site. -And I certainly don't think we deliberately don't want to help people who have to rely on semantic element aware readers)
This isn't the first time these issues have been brought up though. Or even the tenth time. I've seen many small open source projects make accessibility a top priority, actively seeking feedback from visually impaired users, going through multiple prototypes etc. With HN, this just continues to not happen. I could produce a 10-line CSS patch that would solve lots of issues, but at this point I have no reason to believe it would be accepted.
If you offer Dan a 10 line patch which changes nothing in the normal case, but improves things for visually impaired users, I would be surprised if he didn't accept it. I agree there is a low level of visible engagement but I've always found him (and his sidekicks) amenable to reasonable exchanges off this page.
