In terms of reparameterisation, the main opportunity I see here would be a non-centered parameterisation for psi_o. Given that you have a lower bound, you’ll need to construct the parameterisation manually. I’ve previously put together an example over in this post: How to constraint multilevel parameters when using a non-centered parametrisation? - #4 by andrjohns
However, I would also recommend doing some sanity checking of the consistency between the location and scale parameters of the two sets of priors for theta_o. If they’re markedly different, or if they’re strongly informative in different places/directions, then this could also result in some difficult posterior geometry