Ensure Alternate Addresses are Reachable Before Rebalance

There may be a delay between a balance operation making bucket shards live and DNS propagating for that node.  During this window the SDKs are going to have a hard time, causing untold strife for client code.  We should inhibit the balance until all cluster members are reachable.  This will be a raw TCP handshake which takes into account both DNS and L3 networking.