Fix scheduling of jobs to occur only when an ordinary error happens, I mean, while the api is not fixed, do not swallow the exception and do not reschedule