/kind bug

What steps did you take and what happened:

1. Boot a cluster with a given name in the default namespace
2. Wait for that cluster to successfully come up
3. Create a new namespace
4. Boot a cluster with an identical name to the first in the new namespace
5. Cluster does not create a new VPC or any associated resources

What did you expect to happen:
A new VPC, set of machines, and associated resources should have been booted

Anything else you would like to add:

Environment:

• Cluster-api-provider-aws version:
• Kubernetes version: (use kubectl version):
• OS (e.g. from /etc/os-release):