Key sample jobs are very sensitive to configuration

* Hibernate failure
* restart
* rollback
* retry

Users will try similar approaches and use slightly different configuration - options like exception handler, skip limit policy, retry policy, chunk operations all could be set in a way that would break one or more of the samples without a very informative error message. Need to go through a few combinations and see what the error is, and try and make it more obvious what has gone wrong.