When executed with the --remote-ssh-command option, the recover command will activate remote recovery and will use those credentials to connect to the remote server (similar to what the ssh-command configuration option does in the backup phase but in reverse—see the Hot physical backups with Barman recipe). Internally, Barman relies on rsync for this operation.
When performing a full recovery (up to the latest available archived WAL file), Barman recreates the structure of the PGDATA according to the backup. It will then deposit all the needed WAL files in the pg_wal directory.
A careful analysis of the content of the restored PGDATA directory shows that no recovery.conf file is generated by Barman in the case of a full recovery.
It will just simulate a standard crash recovery of PostgreSQL and start replaying the WAL files from the REDO point, contrary to the Recovery of all databases recipe, where recovery.conf was used.
We decided to adopt this strategy in Barman so that we could maintain the same timeline (as a recovery.conf file would start a new era in the cluster's existence), and avoid setting restore_command.