I find the idea pretty wild that we should be able to predict earthquakes within days. My understanding is that earthquakes can be predicted like bridge collapses can be predicted. You can collect signs of a developing fault, but the uncertainty of when it rips is measured in years. And if your model is lacking for the specific instance, you'll misjudge the chances by orders of magnitude.

So I read papers like this one as hopeless fishing expeditions. But if the fishing expedition turns up with a good catch anyway, people will look into it more. Unfortunately I'm not versed enough to asses their catch. Will wait for a geologist friend for context.