Spotlight
- Introduce Fleet as a first-class RunsOn product for GitHub Actions runner scale sets, built around pool-based capacity management instead of one runner per job attempt.
- Add ECR pull-through cache support for runner image pulls, reducing dependency on upstream registry availability and improving cold-start behavior.
CloudFormation
- Disable budget-related resources when the configured daily budget is zero.
- Add an explicit dependency for the Network Firewall endpoint resolver log group to avoid creation-order failures.
Terraform
- Add Terraform support for deploying Fleet pools and their control-plane resources.
- Add reusable control-plane alerting resources for Fleet email notifications.
- Add Terraform controls for creating or referencing ECR pull-through cache rules.
- Allow S3 cache buckets to be configured as account-regional resources.
- Disable budget-related resources when the configured daily budget is zero.
Other fixes
- Include ECS server logs around the job in CLI
--fulldiagnostics archives. - Add Windows S3 cache support for GitHub Actions cache workflows.
- Improve runner resilience when transient GitHub registration authorization failures occur.
- Improve Fleet launch and registration observability, including clearer reporting when spot capacity is disabled.
Release resources
- Upgrade guide: https://runs-on.com/guides/upgrade/
- CloudFormation template: https://runs-on.s3.eu-west-1.amazonaws.com/cloudformation/template-v3.1.0.yaml