I think part of the problem here could be solved if someone was willing to put some work into some scripting... The basic information all exists. But it is not currently exposed in a useful fashion.
My RFCTool (open source, MIT license) automatically translates normative and informational references into citations. So to create a reference to RFC822, I would write:
<norm="RFC822"/>
(Yes this is not XML, I am aware of that, this is markup for use inside either Markdown or Word document source, I regard xml2rfc as being an output format, not an editing format, I mostly support it for input so as to create an editing format)
The tag above causes my tool to note the label starts with RFC and uses the rest of the label to form a URL which is used to attempt to pull the citation from the tools directory. Unlike the XML approach which requires the reference to be stated in three places, I specify it in exactly one.
I don't think I want to change the way I cite numbered RFCs because I sometimes use them to intentionally refer to something that was changed. But we have discussed the idea of short names for RFCs on RFC-interest. We already create such names when we use drafts. Why throw away the information.
Let us say we had used SMTP and SMTP-Message as the short names for RFC-821 and RFC-822. We could then use <norm="RFC-SMTP-Message"/> to refer to the latest version in the document and this could automatically pull up RFC5322 which is the latest version.
What it would take to make that happen is
1) Someone writes a tool to convert existing information into relevant citation files
2) Someone goes through the back catalog that isn't tracked in the database.
One question is whether we actually archive the XML versions of RFCs anywhere, I just had an look and could not see it but I might just have picked RFCs written in nroff.