If a node crashes, messages that initiated the process are not consumed, but redelivered to another node in the cluster or the same node if it becomes operational.
When a message is redelivered, the recovery mechanism starts up.
If there is a process associated with the incoming message in the system, a new process is created and the workflow executed from the first activity.
If a process was already created, the process resumes from the last point the state of the process was persisted.
