I don't think Ctrl+Z adequately simulates a network failure. I recommend you try something else.
If so, is HornetQ capable of resolving such kind of split-brain problem?
firstly you are using replication but the servers are sharing the same journal, this is why you see them using the same node id.
secondly, for any system to handle split brain you either need a single reliable source to control it or a quorum, HornetQ uses a quorum to do this. How well it handles it depends on you intranet infrastructure and the number of servers, 1 live/backup pair certainly wont work.
when you test it you also need to configure TTL etc to manage any network outages you have.