When working around "bad systems that insist on not allowing
updates in an atomic manner", we should not run previous exception
management code that tries to perform atomic move in case of
exception since the dirty non atomic move has already been
performed.