Instance capping is based on the AMI. When using multiple EC2 instances configurations based on the same AMI, the instance capping does not work.

Use cases :

• using the same "raw" image, for instance a basic AWS-provided ubuntu, for multiple configurations done in the init-script or UserData.  
• Having a capped on-demand configuration, plus a spot configuration : X machines on-demand, then spot instances for cost management. The instances would have the same AMI and same configuration. 

Ideally, it would be nice to have an "identifier" just like there's a "description" for instances configurations, and base the capping on this instead of AMI ids that can be used across several configurations.