Jobs can now be automatically retried based on the signal received by the command process that caused it to exit, in addition to the job's exit code.
This is particularly useful in catching terminated agent hosts, such as you'd see when using EC2 Spot Instances:
- label: "Tests" command: "tests.sh" retry: automatic: # Catch cleanly-terminated instances - limit: 2 signal_reason: "agent_stop" # Catch timed-out agents - limit: 2 exit_status: -1 signal_reason: none