Feature: Retrying after network failures

[mzgubin]

Currently if there are any network failures mid-job the job just simply fails. It would be great if we can somehow have it retry from last known copy point. And we can configure it to specify the maximum number of retries.