I'm almost convinced that we're dealing with two different problems here:
- How do we make good suggestions, supplemental links, and other things that actually might make the post better continue to be seen as valued while
- ... we stress the truly ephemeral nature of comments?
It seems to me that, since we keep having this discussion (and sobbing a bit when we query the live DB for the last thousand things flagged), we can treat the first bullet point as the exception to the norm. Most comments are not:
- Essentially suggested edits that weren't made because the editor wanted to collaborate before making changes, or,
- Supplemental links that, while only often tangentially related to the post itself, provides the reader with more to explore, or,
- Counter, contrary or supplemental views that someone managed to articulate in a manner worthy of preservation within the space that we allow for comments
.. why don't we just have a separate spot for that stuff that isn't called comments and see if there's still any angst surrounding automatic self-destruct of stuff that's supposed to be ephemeral to begin with? Because ephemeral stuff should go away, that's why we call it ephemeral.
I'm not pretending to spec out what that should look like here in this answer, I'm answering only to say it looks like we keep hitting our heads against one wall when we could be doing twice the damage to two walls with one head instead of two heads with one wall (or something like that, I'm getting a little lost in that analogy) or well you get it, why aren't we just considering two different use cases since people have clearly demonstrated the need for both since comments became a thing?
Are we looking at talk pages, similar to Wikipedia? Are we looking at more work into hiding older comments by default, but exposing them more to review this time (that doesn't seem like it would scale well), or are we looking at something else entirely?
I'm going to decline this specific suggestion for implementation because it doesn't solve for what worries folks, and I think those worries are valid. But I think the broader conversation here is good, and I'd love to see some proposals that consider two different use cases.