I would recommend creating a backup repo per cluster. You won’t get yourself into any trouble by running the backup on one cluster and then the other, but you won’t get an incremental backup.
When you setup two clusters with XDCR the data will be the same in both clusters , but the sequence numbers and failover table which is used to order events in an individual cluster will be different. All of Couchbase’s backup tools (including cbbackupmgr) rely on the sequence numbers and failover table to know where to resume from when doing an incremental backup.
If you run a few backups on one cluster (in the scenario described) and then point the next backup at the other cluster then cbbackupmgr will end up doing a full backup. This won’t cause any problems for cbbackupmgr, but it will mean that you’re not doing an incremental backup.