High write volume on the dr site witout a clear reason

I've built foundationdb from the github source (master branch) and I'm making some tests on the dr site replication.

I have two fdb clusters a primary and a dr cluster. Each cluster have three nodes. There are three dr_agents running and replicating data from primary to dr.

While there is no any input transaction volume to the primary cluster, 'status details' shows about 25 writes per second tps on the primary cluster and about 3500 writes per second on the dr cluster.

This volume - 3500 - seems to be too high. Limitting the idle write volume by a reasonable value looks like a good idea.