@torkar just a note to say that I accidentally found the post you’re talking about (I think): Use of `sample_prior='only'` in brms
I’m having the same problem with priors that I thought were fairly tight — certainly much tighter than what I’m used to seeing in my field — generating NAs in almost 100% of the pp_checks, and that’s how I came to find your post. Lots of useful stuff there!