[txbm]

Summary

• preserve Dagger infrastructure error statuses for a bounded retry grace period
• continue when the status returns to pending or success
• keep real failure statuses terminal and preserve discovery/wall-clock timeouts

Supersedes #113 now that dagger cloud checks list --watch has been verified to exit during the managed-engine automatic-retry gap.

Validation

• uv run --project github --frozen pytest github/tests (68 passed)
• root Dagger typesafe-daggerverse:github-pytest check (68 passed)
• Ruff check and format passed

[danielgafni]

I don't think this is the right approach because we can't distinguish failing checks from retrying this way.

Wouldn't this approach make waiting for a status report on all runs with at least one failing checks slower by 1 minute? I don't think we want this to happen.

We should push for Dagger Cloud to actually publish a dedicated "retrying" status that we would be able to consume directly.

I don't want to merge work-arounds until we can do that.

[txbm]

I don't think this is the right approach because we can't distinguish failing checks from retrying this way.

Wouldn't this approach make waiting for a status report on all runs with at least one failing checks slower by 1 minute? I don't think we want this to happen.

We should push for Dagger Cloud to actually publish a dedicated "retrying" status that we would be able to consume directly.

I don't want to merge work-arounds until we can do that.

Agreed not worth it, will close and wait for Dagger team