I'm not sure about this 'one wiki per language.' To me, the practice seems to be more on the order of 'one language per written form.'
Identifying a written statement as being, say, English is much different from identifying a spoken statement as being English. Any speaker of English who can read will identify this e-mail as being written in English. An American (or a Briton? I can never remember which does what) might be somewhat surprised at the fact that I spelled "practice" in the first line with a penultimate C rather than S, but this will not prevent the text from being recognized as written English. Even if I were to throw in a bunch of detail about how I went to the dep for a two-four and then went and got the bumf for my carte-soleil, it will be recognized as written English, albeit with a bunch of words thrown in that they don't recognize and will have to have explained.
In the case of Wikipedia, an expression that gives others pause may be defended quite simply. For example, on the fr:wiki, someone changed my discussion of Montreal's "village gai" to read "village gay," which would be standard in France. I simply reverted the change and noted that in Canada, we write "gai" for this meaning, and this was relevant since we were talking of a Canadian toponym.
Note here that it is standard written usage that is at issue: a Quebecer who writes "gai" for "gay" will nevertheless not attempt to write encyclopedia articles as "L'Village gai c'est quet'part où on s'en va pour êt' su'l party pis tenter de pogner, viarge..." Non-standard written forms that attempt to closely mimic English speech will, paradoxically, *not be recognized as written English.'Fi ruyt luy hlis, it is certainly a representation of a form of English, but it will not be viewed as written English, and on the English Wikipedia will be reverted.
What does this all mean? I disagree with Mark where he says that highly nonstandard writing should not be reverted simply because it is English. The medium of en:wiki is not English meaning 'anything produced by an English speaker'; it is English meaning 'English written according to one of the prevailing standards for written English.' That's what's usually meant when "written English" is said. I don't think there's anything wrong with this, as this will be what is, in general, expected by someone who sets out to read something in English. His example is English, but it's not "written English."
If it were to be the case that some form of English were to acquire a written form other than standard English, such that there was a demand for a Wikipedia in that written form, I doubt there would be any difficulty in creating it. AAVE, although a highly distinctive and well-defined form of English, does not, in general, have a written form: a speaker of AAVE who wishes to write in the standard register will use standard English to do so. If transcribed AAVE were to come into use as a common medium of written exchange in the standard register, I would support an AAVE wikipedia.
I'm speaking here, by the way, of lects that are agreed upon to be dialects -- i.e. ones in which the speakers agree that they are speakers of the broader language. This doesn't apply if the speakers believe their language to be distinct -- AFAIK, the case of Scots -- or of independent languages that don't have a long-standing written form or literary translation and are therefore recorded as transcribed speech. In either type of case I would support a wikipedia.
There obviously isn't a one-size-fits-all solution. In general, though, the problem is self-regulating: a speaker of a dialect without an independent written form will generally choose to write in the standard form, or if not the writing will be altered to be in the standard if it cannot be defended contextually; and only written forms of some kind will be able to form independent wikipedias.
Matt