So, you shouldn't probably use the intrepid images at this point, you should use koala. But, if you're already running intrepid (like me), you may run into this stupid problem. If you're using EC2 for temporary workstation / servers, you're probably using freenx. Freenx will mostly work, but will break some random things, including tcl.
It is absolutely critical that you unmount Elastic Block Store volumes before terminating an instance. Otherwise, your data will likely be borked.
An unresolved issue currently - each time I start a new instance from an image with freenx-server, I need to do dpkg-reconfigure freenx-server. Otherwise, it works fine (at least with KDE, which is what I'm using - perhaps foolishly).
You'll almost certainly want to check out the developer's guide.
It looks like the new EBS-backed instances are the way to go, but for now, I use a command like the following to make an AMI from a running machine that I've configured the way I want it:
ec2-bundle-vol -k <private_keyfile> -c <certificate_file> -u <user_id> -p <image_name> ec2-upload-bundle -b <bucket> -m image.manifest.xml -a <access_key> -s <secret_key>
The user_id you can get from "Your Account -> Security Credentials" on the Amazon EC2 website (near the upper right, remove the dashes). The access and secret keys are lower down there as well.
You can then easily register your instance using ElasticFox or the AWS console or the ec2-register tool.