So, after much investigation: turns out my raw data were so corrupt that they contained 7 negative values! So imaginable to me that I didn’t even consider this possibility, but took care of removing NaNs and 0s 
Thus, the “Bad energy issue is gone”, but I’m now in quite a pickle, since the trace stops at ~15%, telling me that the “mass matrix contains zeros on the diagonal”:
I tried reparametrizing the model to a non-centered version (as advised here) and diminished the hyperpriors (as advised here), but the issue is still there.
Any ideas on what to do next? I’m quite lost here…