replication queue scheduling strategy

### Company or project name

_No response_

### Question

I have a clickhouse cluster with 1 shard, 2 replicas, all tables are ReplicatedReplacingMergeTree engine. I add a new replica node into the cluster. On the new node I create tables one by one, when one table's replication queue is 0, the next table will be created.
When replicating a large table A, the replication queue increases rapidly. background_fetches_pool_size is set to 16, and replication_queue shows that there are 16 GET_PART queues for table A. The waiting queues for the other tables have been growing continuously, and the total number of queues reaches tens of thousands.
What is the scheduling strategy for replication queues please? Is there any way to prevent other table replicas from being polled because of a single large table replication?