Automatic retry of incomplete executions

 checks the origin of every incomplete execution when it's created.  automatically retries incomplete executions that have been created because of:

and incomplete executions that were created with the Break error handler with the automatic run completion enabled.

 takes the following steps when  retries incomplete executions:

 schedules automatic retries with the incomplete executions backoff schedule. The backoff schedule prevents the situation when you would get the same error multiple times in a row.

For example, when you get the ConnectionError because the app is unavailable, it might take some time until it's back.  spaces out the retry attempts to get a successful retry even at a later time after the original error.

Other error types usually require changes in the incomplete execution and manual resolving.  doesn't retry these error types automatically by default.

After  schedules the retries,  runs the  again, starting with the module that caused the error.

For each , there is a limit of 3 incomplete execution retries running in parallel. If there are more incomplete executions scheduled from the same ,  retries them in batches of 3 after the previous batch finishes.

In addition, the retry doesn't start when the original  is running already.

The 3 parallel retries limit applies to retries from the same . When  retries incomplete executions from multiple , then each of them has their own limit.

For example:

You have a  that runs for 10 minutes every hour. There was a disruption of a third party service for 5 hours.

If a retry attempt succeeds,  marks the incomplete execution as Resolved and stops retrying.

If all of the retry attempts fail,  marks the incomplete execution as Unresolved. You can then retry the incomplete execution when the app that caused the error is available again or you can resolve the incomplete execution manually.