Gerard,
I disagree with your proposal.
I think some of the proposed criteria are very arbitrary. What is "sufficiently expressive" for a modern encyclopedia? Does that prevent many natural languages (such as Navajo) which don't have words for advanced technology? Wouldn't "insufficiently expressive" languages be perfectly sufficient for the vast majority of concepts, even if they might not have an article on quantum superstring theory?
Further, I've painstakingly followed every thread in this discussion, and I have not seen any strong argument for allowing languages nobody uses natively. Wikimedia wikis exist to make the sum of human knowledge available to everyone, not to practice or preserve languages.
I think the argument that they act as a common language for scholars of the ancient language is not valid; we are not a forum for academic exchange. An English scholar of Ancient Greek can (and probably does) use English in his everyday life, including research and communication. An exception can (and is) made for Wikisource, which exists to collect existing literature, but other projects in dead languages do not serve our mission. A scholar of Proto-Indo-European does not communicate in Proto-Indo-European.
So while I'm open to further debate, I currently disagree with this change.
(As an aside, the vague statement in the policy you point out is only there because you consistently blocked a majority agreement to remove it.)