Target system unavailability period situation…
In order to handle this situation, you can suspend the trigger which will start queueing messages in the trigger client queue. However, you have to be careful about how long this downtime might go, and what would be the maximum number of documents that can get queued (* max size of the message). Queueing messages in trigger can lead to broker down situation if sufficient memory is not allocated.
I assume you are using Broker and not Universal messaging. You can configure JMS Policy based clustering (and choose appropriate algorithm - like Round Robin) which will distribute your messages to all the brokers which are clustered policy based.
Hope this helps…