Thanks for all those interesting suggestions!
As I understand our situation, the aim of our focus shift from EDTF to ISO-8601-2 is not only a way to try to give this standard deeper roots but also a shift from a focus on date representations within the library / archive "culture" to a wider world of digital representation of date-time information.
Keeping this in mind, I consider that we would need a deeper explanation to clarify why "unknown at both ends"
> a construct like ?/? would be explicitly forbidden (as would "/" for open at both ends).
This wider world may include, for example, philosophical ideas. Let's consider a "software" (or schema or some future something ...) requiring date information for (philosophical / ideological) assertions. How would we date the assertion / belief that
"1 + 1 = 2"
(assuming that the characters within this assertion keep today's meaning)? Other examples are welcome, of course. In other words, I suggest to clarify EITHER what "/" and "?/?" (as well as "/?" and "?/") would mean and in which cases these may be useful (in a wider perspective) OR more exactly why we choose to reject them.