While working on the readiness probes we have discovered that the single member health check always allocates a new client. 

Since this is an expensive operation, we can make use of the pooled client (that already has a connection open) and change the endpoints for a brief period of time to the single member we want to check.

This should reduce CEO's and etcd CPU consumption.

any supported version    

always, but technical detail

 na    

CEO creates a new etcd client when it is checking a single member health

CEO should use the existing pooled client to check for single member health