HA-JNDI load balancing only comes into play when talking to it remotely,
or when the HA-JNDI proxy in the server does not belong to the same
cluster as the target HA-JNDI proxy trying to get.

Any other time, HA-JNDI tries to do a lookup locally and if not found, asks
the cluster. Now, out of those responses, HA-JNDI takes the first non null
and non exception one and returns that and hence multiple calls to HA-JNDI
can look as if they were not load balanced.

In order to see some kind of load balancing in this situation, HA-JNDI would
need to somehow be extended to keep track what was the last non local
response returned back to the client and take that potentially into account
when returning the next response.