well i think you’re courting numerical problems by using an improper prior like that. from a practical point of view the point of a prior isn’t only to inject prior information but also to regularize things. in this case some amount of regularization may be important to avoid numerical problems. so if i were you i’d reconsider my prior assumptions.