Recover Docker before macOS tests

[IlyaasK]

summary

• verify the Docker daemon before the macOS test job uses it
• start Colima directly when it is stopped
• force-restart Colima only when Docker is unhealthy or a normal start fails
• recreate the Colima VM if recovery still fails
• wait up to 60 seconds for the daemon to become ready before continuing

why

The self-hosted macOS runner can retain a stopped or broken Colima VM between jobs. The prewarm step requires Docker to start its local registry, so the job currently fails before running tests when that daemon is unavailable.

validation

• parsed .github/workflows/test.yml as YAML
• checked the recovery script with bash -n
• git diff --check
• test-darwin passed end to end, including normal startup, stale-state recovery, prewarm, formatting, and tests

---

Note

Low Risk
CI-only workflow change on macOS runners; no application, auth, or data-path impact.

Overview
Adds an Ensure Docker is running step at the start of the test-darwin job so a stopped or broken Colima VM on the self-hosted macOS runner does not fail the job before prewarm (which needs Docker for the local registry).

The step exits early if docker info succeeds; otherwise it tries colima start, escalates to force-stop/restart and then colima delete if needed, and polls up to ~60s for the daemon before failing with docker info.

^{Reviewed by Cursor Bugbot for commit ce45c8a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[sjmiller609]

thanks, this is very helpful but just one thing

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit e36e29d. Configure here.}