As long as the mail files are based on the mail9.ntf template just replacing the one that is there will get updated the next time the design task runs or you can run the nconvert command http://www-01.ibm.com/support/docview.wss?uid=swg21304939.   As for the pernames.ntf there is likely one in the client update and once the client starts the local templates should get upgraded based on the templatesetup  line in the notes.ini that should be a different number.   If not just set this line to 48 or anything and the running of the notes client wil do the template updates.

Is your question about how do you provision multiple language versions of mail9.ntf and pernames.ntf on the same server?  On the Design tab under Database Properties, do you have the appropriate language selected in the section for Multilingual Options?  

Exactly! In every mail file either English, German, French or Italian is selected in the section for Multilingual Options in the Database Properties. Our templates are multilingual since we installed the Language Packs.

The same applies to the users roaming databases (names.nsf).

In the past we've installed additional Language Packs on the Domino server.  In doing so, this would, in an additive sense, add the various multi-lingual design elements to the existing mailx.x mail template.  We found, at least in R8.x, that by having too many (we have 10!)  languages installed in this fashion caused the mail file design itself to grow too large (too many elements) and we could no longer modify the [ACL Default] preferences once that occurred.  In addition, creating mail files based upon language AND also using the Load Convert -gLANGUAGE option took far longer for the server to process due to the number of design elements that had to be exempted during db creation or design updating.  When we went to R9.x, we Installed the Language Pack one language at a time against a default mail9.ntf (English) mail file, and then made a copy giving it a file and template name related to the Language to prevent this.   Kudos to IBM for issuing the new language specific template.  I have not looked at them yet, but certainly do hope they all have a different replica ID.

The solution was to merge different languages into one template with the program "mlmerge.exe"