memory leak in backup server

We have a MariaDB instance dedicated to backups only. It is only receiving replication traffic and taking backups with mariabackup.
This instance consumes all available memory over time, until OOM killer intervenes.
The server is an EC2 instance with 186G of RAM, and despite having set innodb_buffer_pool_size to 70G, it consumes all memory in about 7-8 days and we have to restart it to avoid OOM.

# ps -eAo pid,rss,args | sort -nrk 2 | more

 238426 173594784 /usr/local/mysql/bin/mariadbd

I suspect the leak is with optimistic parallel replication.
Here are the relevant settings in use:

slave_parallel_max_queued=94371840

slave_parallel_mode='optimistic'

slave_parallel_threads=24

innodb_buffer_pool_size=75161927680