I have a very strange case. We have an OLTP set up with AG and transnational replication. We do around 3K transaction per second when it is not really busy. We have noticed replication latency between Publisher and Distributor when we have long running select queries. Has anyone seen this scenario? Is there anything that can be done about it?