Someone who speaks Ursari, the dialect of most Roma in Rep. Moldova,
will have a difficult time understanding anything spoken or written in
Vlax, and it may be impossible.
This would be like you having to read all WP articles in Aromanian, I think, or maybe worse.
According to Ethnologue, the different varieties of Romani are sufficiently different to be considered separate languages, even though they are part of the same Romani macrolanguage (code ROM - I have no idea why Vlax was allocated RMY).
However, the following source claims that Romani activists from all around the world were able to understand each other without interpretation:
http://www.b-info.com/tools/miva/newsview.mv?url=news/2000-07/text/jul28d.rf...
Regarding script, while it is certainly true that Devanagari is
obviously crafted for Indic languages and thus fits naturally to the
phonology of many (though not all) Romani varieties, as far as I know the movement to use it has very little actual support among the Roma community, or rather, the *literate* Roma community (important distinction).
Nearly all literate Roma use Latin AFAIK (with some using Cyrillic although not nearly as much as earlier years), and it doesn't seem possible that there is a single person literate in Devanagri but not Latin for Romani.
When I first launched the proposal for the Romani Wikipedia, the intention was for it to be Latin only. Personally, I believe that it would be best for the language if Latin were to be the standard alphabet (and in many ways, it already is). However, I noticed that Desiphral started to use Devanagari. Initially I asked him why he wrote in Devanagari, since it was not used at all on the Internet, and he explained that it's more natural, and that the ancestors of the Romani wrote in this way, and hence it best represents the language. He also said that another contributor, Naayram, also preferred this alphabet.
Along these lines I think it would be a bit funny to make a Vlax Wikipedia entirely biscriptal with both scripts on equal footing. I think it would be a good compromise to have Latin as default, and if people wanted to, they can click a button to convert. I also think it's a bit strange to have both on interwikis, but whatever.
Yes, I think the best solution for conversion would be the Serbian-style model, where Latin (in the sr: case, Cyrillic) is the main script, and content is stored in Latin, with the ability to convert to Devanagari. I think a true biscriptal solution, where there is a Devanagari database as well may be too hard to implement and may not have the necessary demand from Romani speakers.