Maxscale doesn't allow granular control over repl channels

I have a use case for a downstream cluster that maxscale consistently breaks my intended replication flow.

We have an external master sending a specific schema/table list that should go to the local cluster's master always, with the other node being a full slave of the "Master, Running, Slave of External Server".

External Master
│
│ host_master (filtered schemas)
▼
[Active Master] ──── '' (no filters) ────► [Active Slave]

On failover, the promoted node already has host_master from my.cnf to define those outside schemas. I intended to have the a node rejoin on channel '' but from code I see

auto_rejoin selects the channel to redirect by taking m_slave_status[0], the first channel in SHOW ALL SLAVES STATUS order and since host_master is defined in my.cnf it starts first on boot and is always index 0. MaxScale redirects it to the new cluster master, leaving the returning node replicating via a filtered channel — silently dropping writes to non-filtered schemas. MaxScale reports the node as Slave, Running with no error. Slave status ends up looking like

[13:22:45] root@localhost:(none) > show all slaves status\G

*************************** 1. row ***************************

               Connection_name: host_master

               Slave_SQL_State: Slave has read all relay log; waiting for more updates

                Slave_IO_State: Waiting for master to send event

                   Master_Host: Internal_master

                   Master_User: repl

                   Master_Port: 3306

                 Connect_Retry: 60

               Master_Log_File: mysql-bin.001043

           Read_Master_Log_Pos: 11781707

                Relay_Log_File: relay-bin-host_master.000002

                 Relay_Log_Pos: 7424

         Relay_Master_Log_File: mysql-bin.001043

              Slave_IO_Running: Yes

             Slave_SQL_Running: Yes

               Replicate_Do_DB: db_domain,db_feedback,db_kpi,db_non_shared,db_program_config,db_scoring,db_service_catalog,db_user,db_user_activity,db_unbounce,db_billing,db_email,sg

           Replicate_Ignore_DB:

            Replicate_Do_Table:

        Replicate_Ignore_Table: db_non_shared.email_track,db_email.recipient_status_queue,db_non_shared.email_bounce,db_user_activity.tbl_username_log_archive,db_user.tbl_bill_html_archive,db_email.queue_archive,db_email.user_email_status_archive,db_user.tbl_email_queue_archive,db_email.recipient_status_queue_archive,db_email.all_queue

There is no configuration option to tell mariadbmon which channel name to use for internal cluster replication in my case.

I built maxscale with a change to allow a rejoin_channel monitor variable and am able to set that to 'int_master' via a code change to cluster_manipulation.cc and am able to properly let maxscale maintain my desired architecture. More than will to share those full changes

I set this as a bug because it can cause dataloss in my setup but it does seem more on the feature side.

Thanks,
Bryan