This issue is rather interesting from the Bayesian point of view: the priors for superluminal neutrinos are so low, the evidence for them, even when taken all at once, like in the paper you cite, is still not compelling enough to assign high credence to the model.

Basically agreement here although I think that some degree of disagreement may show up when one unpacks "high credence." It would seem to me that the evidence is sufficient enough to merit detailed investigation, especially because many of the obvious things to do are experiments which will provide other interesting data. At minimum, this is another reason for careful attention to the neutrino distribution the next time there's a supernova as close as SN 1987A.

As it currently stands, Ehrlich is rather biased, having written several pro-FTL neutrino papers on the issue, including one withdrawn from arXiv after the issues with OPERA came to light, and so his writings do not increase the credence of the FTL side much.

I'm not sure I agree with this. In this context, most of the paper's analysis is pretty straightforward so I don't think his biases are that relevant to evaluating it. I'd still put a higher probability on some sort of systematic error going on here than on FTL neutrinos (by a wide margin), but that's for essentially the same reasons you mentioned (the hypothesis having a verylow prior) together with the empirical problems like SN 1987A.