We decided HIVE-2837 was going to have too high of a risk:reward ratio. However, we believe there's value in adding code to discover and in some way report such leaks as long as that code is passive (does not effect changes in the cluster or cloud) and inexpensive (does not incur (significant) extra SDK or etcd lookups).

In the context of HIVE-2837, the resources which would be candidates for this detection were:

• VPCEndpoint
• HostedZone

We believe the former should be doable, as we're already querying for all VPCEndpoints.

However, the latter is disqualified as it would entail (numConfiguredVPCs-1) additional SDK calls to look for the redundant HZs in the other VPCs.

cc drow.openshift.srep jbranham.openshift cbusse.openshift