If the system is up and running and you reboot one DR node at a time and allow it back in the cluster and fulling in sync and then reboot the next DR in the cluster and let it fully sync after reboot and when you get to the last DR node in the cluster the DA service stops when the DR goes down.
Timeline:
DA up
DR1 reboot and fully sync - DA up
DR2 reboot and fully sync - DA up
DR3 reboot - DA down
DR3 fully sync - You have to restart the DA.