Stacklet images regenerate ssh host key, making status check fail

So, when installing from a Stacklet image in Cloudmin (maybe not all of them, but it seems like it would be a good practice and maybe we ought to be doing it in our images, too), the status seems to always end up with "No SSH", not because startup failed, but because the host key gets regenerated during reboot, which leads to an "invalid key error" during the status check.

Fixing the problem is as easy as removing the offending host key from the /root/.ssh/known_hosts file, but it is not at all apparent from the error that that's what needs to happen, and it feels very clumsy, and could be confusing.

So:

A better error would be good.

But, an automatic deletion of the host key before the first boot, so that the error doesn't happen during status check, would be best.

The Stacklet image I saw this problem on is the CentOS 5.11 KVM image. It's already downloaded on srv1 if you wanna try it.