If an AMI referenced in a cloud is deleted but a reference to it is retained in the cloud config, following behavior is observed
- A build requesting the orphaned cloud's label will wait in the queue
- Jenkins will flood the associated amazon account with a sufficient number of API calls to trigger rate limiting on the account, effectively taking it out of service
- The frequency of API calls never backs off.
This is a very dangerous bug, and can take down your jenkins server and every server on the same AWS account that requires activity on the AWS API. Some sort of exponential backoff would be a far better behavior.