Currently, errors that occur in the workload are hard to investigate because of a lack of observability. Logs need to be improved and a backup should be created on failure:

• Implement backup functionality when an error terminates the workload. Save the backup in the artifacts directory.
• Improve transaction logging to be serialized and show worker ids. When a transaction fails, log concurrent, in-progress transactions.
• Improve logging format - using json would make it easier to inspect programatically