Is it taken as a serious reconstruction, like Proto-Indo European but for folklore? An interesting mental exercise? A kind of loopy concept brought on by reading Clan of the Cave Bear too many times?
I've just read this "Paleolithic Turn" essay and am kind of intrigued.
Not a folklorist or neolithic archaeologist, but as a linguist, I've got to be incredibly skeptical, based on this review. Reconstruction necessarily depends on reconstructions after a fairly short time-depth, and the comparative mythology done by, say, Calvert Watkins, necessarily depends very closely on linguistic reconstruction. That's not a workable process at the time-depths Witzel is talking about--not because historical linguistics is 'unfashionable' (though it is) but because actual historical linguists generally agree that "superfamilies" are built on very shaky evidence. Nostratic, for instance, isn't even built on the same method as Indo-European. While Indo-Europeanists look for regular sound correspondences (e.g. English initial /t/ and Latin initial /d/ match up regularly in a huge corpus of words, and where they don't, we can generally identify a condition responsible for the mismatch), Nostraticists, Altaicists, et al. work with the 'method' of mass comparison, where you pick a list of word meanings, find words with those meanings in the families you're interested in, and check to see how similar they look. There's a high level of rigor involved in the comparative method that is absolutely lacking in the work that claims to reconstruct linguistic families that are older than ~10000 years. This isn't, as Witzel would have it, a "superficial" rejection, by the way.
I just read the article, and Witzel isn't even telling a just-so story -- he's telling an "it's so whether it's so or not" story. Take a look at this statement, as just one example among many:
Apart from the feature of drift, certain motifs that widely appear in Laurasian and non-Laurasian mythologies may belong to an older stratum of pre-Laurasian (and pre-Gondwana) mythology.
This is in a section where he's discussing counterchecks that are supposed to show where the geographical boundaries of Laurasian mythology lie -- and right in that section, he says (paraphrased), "but if those other mythologies do contain elements in common with Laurasian mythology, then that must be evidence of some even older shared mythology."
Data-driven science this is not.
That said, this isn't very different from what Fraser and Campbell were trying to do, so it's not an unprecedented type of overreaching (for whatever that's worth).