Oh yes indeed, I'm well aware of this class and even used it in the
lingvoj.org ontology. See
http://www.lingvoj.org/ontology.
I'll be back to this one, but quickly said my concern here is not the absence of a range class, but the absence of an authoritative list of instances, e.g., lack of an authoritative list of URIS for ISO-639 codes. This is a quite different animal. I've read in this list that LoC was about to release such a list of URIs, hopefully they will use the above class. The question is to know if DC should recommend at some point a specific publication in the Linked Data cloud of instances of this class, or the other way round, infer that any value used of dct:language belongs to this class.
This is a general issue with rdfs:range declarations, which are used by some systems (controlled data bases) as a constraint allowing only values declared in a trustable vocabulary as being instances of the range class, and by other systems (open world reasoners) as a classifier rule. Rarely both uses are considered together.
Bernard