Without known keys, how can you be sure that the server Veeam is contacting is really the one that you think it is? I know the repository should be physically secured, but that might not always be the case, let's assume it isn't "100%" physically secure in this case. By the way, let's also assume that the repository's disks are encrypted, so the attacker shouldn't be able to get to the original data.
What if an attacker manages to get to the remote repository, disconnects it and instead puts his own Linux machine there, with SSH offering logins with the same account as Veeam is configured to use; Can we have a situation where Veeam happily sends the backups to this rogue repository (and the attacker thereby getting copies of the virtual machines)?
I think it depends on a few things, for example whether Veeam refuses to proceed with the backup if the remote repository isn't in the same state that it expects (e.g. has the same backup files on it already, as opposed to being empty).
Regardless, this is one type of scenario where keys would most likely be of great help to mitigate such an attack.
EDIT: On a related note, this touches on the feature request of encrypted backups..
