After the job has been running for about 300 seconds, I get following exception:

12:04:35,504 WARN [com.arjuna.ats.arjuna] (Transaction Reaper) ARJUNA012117: TransactionReaper::check timeout for TX 0:ffffc0a80052:-13db475b:54268a64:1f in state RUN
12:04:35,507 WARN [com.arjuna.ats.arjuna] (Transaction Reaper Worker 0) ARJUNA012095: Abort of action id 0:ffffc0a80052:-13db475b:54268a64:1f invoked while multiple threads active within it.
12:04:35,508 WARN [com.arjuna.ats.arjuna] (Transaction Reaper Worker 0) ARJUNA012108: CheckedAction::check - atomic action 0:ffffc0a80052:-13db475b:54268a64:1f aborting with 1 threads active!
12:04:35,509 WARN [com.arjuna.ats.arjuna] (Transaction Reaper Worker 0) ARJUNA012121: TransactionReaper::doCancellations worker Thread[Transaction Reaper Worker 0,5,main] successfully canceled TX 0:ffffc0a80052:-13db475b:54268a64:1f

Since the default transaction timeout settings in wildfly is 300 seconds it seems that there is one long running transaction executing that starts when my batch job is started and I don't see where this transaction comes from and why it is started. Also, this only seems to happen with a partitioned step.