Autoscaling: When one changes the AMI, then old stopped machines are not removed

[sanderegg]

When changing the configuration of autoscaling, the service automatically removes old EC2s if the pre-pulled images listing changes.
This mechanism does not work if one changes the AMI-IDs, then it will not update the warm buffers.

The same is true if one changes the custom boot scripts.

This usually results in "lost" warm buffers which could become costly.