Confusing cluster launch behavior for failed instances

[hustonhoburg]

I launched a cluster where some of the instances return this error:

 InsufficientInstanceCapacity => We currently do not have sufficient 
 m3.xlarge capacity in the Availability Zone you requested (us-east-1d). 
 Our system will be working on provisioning additional capacity. 
 You can currently get m3.xlarge capacity by not specifying 
 an Availability Zone in your request or choosing us-east-1a, us-east-1b.

The launch continues for the provisioned machines, but fails to launch those that threw that error. This seems like a case of "undefined behavior". The end result isn't completely unexpected given the logged errors. However, my expectation was that if I launch a cluster, then the entire cluster will fail or succeed. My expectation wasn't partial success. It seems to me like maybe the entire launch should fail and revert if some of the machines fail to launch.