In particular, there seems to me no reason why alternative implementations
can't be added to the git archive, perhaps in subdirectories.
I'm willing to do that, but I want us to have a convention that makes it absolutely clear that the code might not have come from the author(s) of the SRFI document and reference implementation. Authors work hard on what they submit, and some will be rightfully worried about their reputation if other people's code is added after the fact.
How about "contrib/" with a "README" file explaining who authored each file in that directory?