Fault-tolerant CI.
If your GitLab Runner is killed - maybe the configuration needed to be changed, or an autoscaler decided to consolidate the number of runners - then all running jobs are killed with a less-than-useful message:
The job restarter listens for jobs that failed due to the GitLab Runner stopping, and can retry them.
The restarter can be configured to limit the number of times jobs can be automatically retried, preventing a job that crashes the runner from continuously bringing down the infrastructure.
In addition, the restarter can avoid retrying long-running jobs, ensuring that using large amounts of resources only occurs when directly triggered by a user.