But we are not talking about existing mbox-reading software (which is
either tailored to some particular mbox format, or else tries to cover the
most general case it can imagine).
No, that's not what "we" are talking about. And I don't claim to know
what _you_ are talking about.
I don't believe that the creation of a new content-type is likely to
result in a lot of new mbox readers or writers. And I do believe that
whatever new software is written to read or write mbox files will be
similar to existing software that reads and writes mbox files. If most
existing software doesn't need knobs to deal with format variation, I
don't suppose that most new software will need it either.