Having looked at a number of crowdsourcing options over the years, I've recently been more attentive to how a Serres translation might be accomplished via Amazon's Mechanical Turk.
I'm certain it would raise the ire of many academics but if the choice is between no translation of any sort for a particular book or a less-than-perfect translation, I'd lean toward having something that can be refined and improved rather than having nothing.
The structure of Amazon's Mechanical Turk is that you take a big project (like a book with 100,000 words) and the Mechanical Turk breaks it up into little bits that any number of people can work on (say a sentence or two of translation at a time). Then the material is reassembled and you can review it, etc.
I have no idea about relative costs, quality, feasibility, etc., but rather than waiting around for some of Serres's work to be translated, it may be a possible angle to consider for decreasing the turnaround time from years/decades to weeks/months.