The documentation for the replication feature describes the new JDBC connection attribute "slaveHost" this way:
"Specifies the system that will serve as the slave for database replication." 
This is true (if system == network host) when used on the master side in a replication scenario.
However, when used on the slave side in conjunction with the "startSlave=true" attribute, the slaveHost attribute specifies the network interface on which Derby (the slave database) will listen for connections from the master. This is not clear from the description of the attribute in the reference guide, which also says
"It may be specified in conjunction with the startSlave=true attribute; if it is not, the default value is localhost."
If the slave is started on a host that is separate from the master host, then the slave must listen on an interface that is reachable from the master, e.g "slaveHost=0.0.0.0" (all network interfaces, IPv4). This is analogous to the Network Server's -h option and the derby.drda.host property, and should be clarified in the documentation.