Open
Description
In which case the audit's behavior is not validating when testing in dev. A problematic workflow may look like following:
- The user adds a new audit to an existing model and references a non-existing projection in its query
- SQLMesh detects a metadata change and applies the plan without evaluating anything
- The user deploys to production where the audit fails during the next
sqlmesh run
execution
We need to come up with a way to verify added audits without evaluating associated models when applying the plan in a dev environment.