Do we really need the complete-upload problem type?

[LPardue]

In a PR discussion, a comment was raised:

I was mainly thinking that 410 and 404 make almost no difference to the client, and it doesn't make much sense for a server to forget the final response but remember an upload happened to a particular temporary resource.

If we decide to use a 0-length append to retrieve the final response, every other error case is covered by mismatching-offset or inconsistent-length so we don't need completed-upload problem type.

Originally posted by @guoye-zhang in https://github.com/httpwg/http-extensions/pull/3252#discussion_r2366317432

[Acconut]

This seems very related to https://github.com/httpwg/http-extensions/issues/3270, which is a proposal to allow 0-length PATCH requests after the upload is completed. And then:

We can eliminate the "completed-upload" problem type altogether if we allow this.