Most languages with cases seem to be either gaining or losing them diachronically (The Indo-European languages are an example of the latter, and the Uralic languages of the former). Manchu and Xibe (a Manchu dialect spoken by the descendants of a Manchurian guard unit posted away from their kinsmen) have the same number of "major" (grammatical?) cases. Specifically, the Manchu dative-locative corresponds to the Xibe locative, the Manchu ablative to the Xibe dative, and the Manchu prolative to the Xibe ablative. For background, see the Wikipedia article on Manchu cases.
How common is this reassignment of case usages without absolute gain or loss?