When both HA instances negotiate their state during connect and when both HA instances are in state STANDALONE, one of them is shut down with an exception and the other continues operation. Both HA instances in state STANDALONE is called "partinioned" (or "split brain") because both HA instances operate separately (partinioned) without a connection to each other.
A partinioned SwiftMQ HA Router has 2 different stores (with Replicated File Store) or use the same shared store. The latter case will certainly throw more exceptions from the store itself.
The reasons why a SwiftMQ HA Router can become partinioned are:
If you use Replicated File Store, you must decide which HA instance you want to use as ACTIVE/STANDALONE instance. If it is the HA instance which was automatically shut down, stop the other HA instance immediately. Delete file haspool/<instance>/ha.state of the HA instance which should operate as STANDBY. Then start both HA instances.
If you use a Shared File Store or Shared JDBC Store, you will have an inconsistent store. Shut down both HA instances, delete file haspool/<instance>/ha.state of both HA instances and delete the store content. Then restart both HA instances.