Automatic Retry for Failed Deployments / Integration Tests

Standard procedure when a pipeline fails in a CSP deployment or integration test step is to retry the job. The purpose of this is to determine if the failure is a temporary failure in the infrastructure, or a more lasting failure (possibly related to the code change itself).

These retry operations can be automated by GitLab itself, reducing the need for the submitter to notice and act on them.

Also, during the tagging week, the script that generates the Tagging Notes retries any job on the latest pipeline that has a failure. This logic is more precise -- it only retries the deployment and integration test steps, avoiding wasting time re-invoking compile or scanning operations.