Currently, the remote-cache-container tries to auto-configure the SNI host name for the remote servers of a cluster via the source address of the socket bindings - and only when possible, is hostname validation enabled.  In retrospect, this would only ever work if all cluster members were colocated.

Given that a SNI host names need not correspond to real hostnames, we should decouple this configuration from the outbound-socket-binding and instead expose a configuration attribute per cluster, and only if set, auto-enable hostname validation.