It looks like the major contribution of this paper is in a new signal processing technique. They have something they call the 'chained power spectrum' which ge

It looks like the major contribution of this paper is in a new signal processing technique. They have something they call the 'chained power spectrum' which gets used to boost a 4 sigma signal into an 11 sigma one. I don't have all morning to dig into their math but the results are going to live and die on their modeling assumptions. (However I do note that in 4.2 they admit that the choice of "PPP" (Eq 22, section 3.5, all references to the arxiv version) fails in simulated data and of the alternatives that don't fail that particular test, they went with "PQP" instead of the more conservative "QQP" for reasons that I have trouble as interpreting as 'it made the results flashier'. But I'm just some weirdo on the internet, and it's not my field, and I've got to get back to work).