If you are ok with gaps in message delivery and you are using persistent messages then yes.

A network is always store and forward, so the state is not replicated, it just moves between brokers on demand.

If a broker goes down, any messages on the broker (at the time that it stopps) will be unavailable until that broker restarts.

From an application perspective these messages will appear as missing and other messages may be received before them (order cannot be guaranteed in a complex network scenario)

A typical deployment uses master/slave pairs for netwok nodes such that a slave can take over on demand and make pending messages available very quickly.

If you are using fire and forget non persistent messages, then inflight messages will be replayed, but pending messages will be lost if a broker dies before it can dispatch them. ie: normal non-persistent failover semantics