Description:

Failed prebuilds currently linger indefinitely, which can cause resource clutter and cause confusion. While these failed prebuilds are useful for debugging, they should be cleaned up automatically after a reasonable period.

As a possible solution, we could extend the existing expiration policy to include failed prebuilds, potentially with a different retention period than successful prebuilds. This would help keep the system clean while still allowing enough time to investigate failures.

• Implement automatic cleanup of failed prebuilds
• Update documentation and monitoring accordingly.