eks deploy should be robust to eventual consistency

Should avoid errors like the following

```
ERRO[2021-02-08T21:01:38-04:00] Error retrieving detailed about the instances name=kubergrunt
ERRO[2021-02-08T21:01:38-04:00] Undo by terminating all the new instances and trying again name=kubergrunt
ERROR: InvalidInstanceID.NotFound: The instance ID 'i-028ec69838e751d18' does not exist
status code: 400, request id: 9c623f38-d818-43f1-a335-d370a2a5ad76
```

Start by tracing the EKS deploy path that retrieves detailed instance information, using the reported InvalidInstanceID.NotFound log as the reproduction symptom. Check how deployment handles eventual consistency and verify that a delayed instance lookup no longer causes the deployment to fail during rollback.

Written by the indexing model from the issue text.