While developing and testing the Quay/Clair integration, we should be able to create faults on demand to actually test the client's errors handling.

We tend to fix Clair's behavior as quickly as possible, which makes testing Quay's usage against Clair errors difficult.

Open questions:

• Should this live in Quay's client or Clair's API server?
• Should this be some sort of feature flag or a magic layer?