Customers have gotten into trouble using the Amazon domain names for their servers on EC2. When a machine does down in Amazon you cannot control the domain name of the machine that you will get as a replacement. This can be especially problematic when try to replace a config server that has crashed. The level higher level of naming (either CNAMES or Amazon route 53) should be getting used to avoid issue.

This is probably documented elsewhere, but deserves to be explicit on the EC2 page as well.

Asya and/or Jason Z. might be able to provide better details.