restarting replica that has memory engine tables BREAKS replication

# setup at least a 2 node cluster w/ replication & gtid_strict_mode=ON

bash cs_package_manager.sh install enterprise 10.6.14-9 --nodes 172.31.47.230,172.31.40.126

# master

mariadb -e "create database test"

mariadb test -e "create table t1 (id int) engine=aria"

mariadb test -e "insert into t1 values (1)"

mariadb test -e "select * from t1"

mariadb -e "show variables like '%gtid_binlog%'"

# replica  - notice good

mariadb -e "show slave status\G" | grep -E "_Running|_Error|Gtid_IO_Pos"

# master

systemctl restart mariadb

mariadb -e "show variables like '%gtid_binlog%'"

# replica  - notice good

mariadb -e "show slave status\G" | grep -E "_Running|_Error|Gtid_IO_Pos"

# master

mariadb test -e "select * from t1"                # expected to be empty

mariadb -e "show variables like '%gtid_binlog%'"  # expected to be incremented by 1

# replica  - notice good

mariadb -e "show slave status\G" | grep -E "_Running|_Error|Gtid_IO_Pos"

# NOW BREAK REPLICATION - this is the bug

# replica

systemctl restart mariadb  

mariadb -e "show slave status\G" | grep -E "_Running|_Error|Gtid_IO_Pos" # appears good

mariadb -e "show variables like '%gtid_binlog%'" # appears good

mariadb test -e "select * from t1"  

mariadb -e "show variables like '%gtid_binlog%'" # NOW BAD gtid_binlog_pos=0-2-18

# master

mariadb -e "create database test2"

mariadb test2 -e "create table t2 (id int) engine=aria"

# replica

mariadb -e "show slave status\G" | grep -E "_Running|_Error|Gtid_IO_Pos" # replication now shows bad