While distributed work with rcs generally works fine, the naive approach has the disadvantage that the whole history of that file is transmitted. While this usually works fine (it's only the full history of a file, not of the whole repository), some files have a history with a huge number of revisions.
What I do here, is to create an rcs file with shortened history and a tagged revision that is the same as in the original file (I use the script rcsshort for this). Now, in the computation of the hashes to identify equal revisions, I ignore all revisions that are not descendants of that tagged base revision. Of course, all hash values change, but since they're never persisted but only live for a run of rcscjoin, this is fine.