$ 0 0 The only (obvious) solution I can think of is:Take a backup of all tables involved before starting the jobRestore the backup when the job fails.