Rounds proceeds even previous one have failed #61
Description
Ideally, we would like training jobs to raise an error and exit whenever an error occurs. However, when we sometimes have failed jobs due to network/device issues, the app would still proceed to attempt the next round.