According to Snowflake documentation and the community discussion with strong consensus (100% agreement and multiple upvoted comments), when compute resources for a virtual warehouse fail to provision during startup, Snowflake automatically attempts to repair the failed resources. This is the standard behavior documented in Snowflake's official user guide for warehouses and tasks. The other options are incorrect: 'Restart' would imply stopping and starting again rather than fixing the provisioning issue, 'Queue' refers to task scheduling rather than resource provisioning, and 'Provision' is what initially failed and needs repair.