It appears that the superpmi replay job (1, 2) fails if the only problem is SuperPMI MISSING data due to drift between JIT and SPMI collections and the JIT/EE interface calls made.
We shouldn't fail the job (or report an error) if that occurs; it should be ignored.