The question is whether a newly accepted language version can really grow to a respectable Wikipedia. It is not really the point whether there are native speakers, how many speakers there are, whether the language is taught at an university. This only means trying to find criteria that will tell us what to expect from a language version. Nowadays, with 2-6 years of experience, we see that - among the planned languages, only Esperanto is doing well, as interlinguistics could have told us before - among the "ancient" languages, Latin is doing the best, but does not impress. An average la.WP article has only 929 bytes, compared to 3075 bytes in the Alemannic-WP. Many articles in la.WP are no real articles but rather data base entries. Even poorer is the status of Anglo-Saxon-WP or Pali-WP. - among the dialects or small languages of West-Germanic origin, the situation is very different. My study up to now shows that only Frisian and Luxemburgian do quite well, presenting to their speakers a somewhat decent encyclopedia about regional subjects. Bavarian-WP is mostly a joke, often trying to describe things in an amusing way.
After a survey of user communities, it seems to me that a working WP needs at least 15 steady Wikipedians, who speak the language at level N (native) or 4 (or 3 at least). This is true for the Frisian WP, and even more for the Luxemburgian. These two West Germanic varieties do have language status, but this seems to be less important, as eo.WP is quite okay, although Esperanto hardly has status (but more than 60 registered users with level N or 4).
The dedication to a language also seems to be important; it often lacks when a regional language is not really an unified language and when it does not differ significantly from the roof language (look at the Zealandic or Ripuarian WP). It is not enough an "Abstandsprache", a sociolinguist would say.
On the other hand, Wikipedias with less than ten collaborators proficient in the language do not really make a chance, like Upper Sorbian, or the Phantom Wikipedia of Volapük (with two "speakers" of level 2 and three "speakers" of level 1, but with a cunning bot programmer and 112.000 geographical stubs).
I am not sure about the right policy. I find it legitimate if a regional language tries to improve a language with its speaker's community. But it might be a good idea not to encourage groups of five or seven persons who will not even have translated all of the MediaWiki after two years.
Ziko