When setting up a trust between JIRA and a Confluence, one has to specify each individual IP address for each node apart of the cluster separated by comma; or use wildcards. As when the request is coming through the LB, one does not know which specific node will be handling that request. Hence, if the request is delivered to a confluence node that has not been added to the trusted application list in JIRA, that particular user will not be allowed to see any information coming from JIRA (for example the jiraissue macro)

To improve this I would suggest to use delegation approach. When a trust is established with one node, as each node is aware of the others, a message should be displayed suggesting the other nodes IP addresses to be inserted into JIRA admin section. Alternatively, if possible, this information could be delegated and automatically inserted into the list of trusted applications in JIRA via some remote mechanism.