I was able to recover this by removing the cloud provider section of the cluster.yml, rebooting one of the control planes, and then deleting the "new" etcd and control plane nodes from Rancher. There were duplicates, one of ip-xxx-xxx-xxx and one of ip-xxx-xxx-xxx-xxx.region.compute.internal, I deleted the latter. After about 10 minutes the control plane and etcd nodes cleared up.
I would still like to get the cloud provider working, it seems like my only option is to change the host names of all the of the nodes to the correct format? Is that possible to do in a running cluster?