Nixops deletes GCE resources in the wrong order sometimes

[3noch]

I only have one example. For a GCE deployment, if you define a health check and target pool that uses it and then remove both of them, nixops -k will try to delete the health check first and fail because it's being used by the target pool. Deleting the target pool manually works because nixops -k can successfully delete the health check. It then reports that it can't delete the target pool because it's already gone. :)