Error 1827 on MySQL Replication slaves when MySQL Router creates a user

I'm running MySQL 8.0.37 on three different servers, configured as a master with two slaves. All of the versions are exactly the same. My applications connect to the cluster via [MySQL's Community Router](https://dev.mysql.com/doc/mysql-router/8.0/en/mysql-router-installation-docker.html) . My replication system uses GTID replication, so that I can manage the cluster with MySQL Shell. Every once in a while after the pods are recreated I get Error 1827 on the two slaves about creating a "router" username/password. This causes replication to stop (but doesn't prevent select requests from continuing to hit that slave, causing old data to get returned). My current workaround is to skip those transactions individually on each slave using the SET GTID_NEXT command. I've tried setting print_identified_with_as_hex=1 which doesn't prevent the error either. It seems that the query being attempted works on the master but not on any of the slaves. It also seems to be transferring passwords in plaintext, which probably violates some security threshold. How do I prevent this from happening? Is it a problem with my Replication settings, or my MySQL Router settings?