What this bugfix is missing is inserting an "SET @@SESSION.GTID_NEXT= 'AUTOMATIC'" into the master binlog stream after gtid_deployment_step is flipped from OFF to ON on a master that's taking write workload.
The lack of the "reset" GTID_NEXT statement manifests in gtid_next on the slave executor THD being "undefined" and stopping the thread. That said, "start slave;" fixes the issue and no statements are dropped.
What this bugfix is missing is inserting an "SET @@SESSION. GTID_NEXT= 'AUTOMATIC'" into the master binlog stream after gtid_deployment _step is flipped from OFF to ON on a master that's taking write workload.
The lack of the "reset" GTID_NEXT statement manifests in gtid_next on the slave executor THD being "undefined" and stopping the thread. That said, "start slave;" fixes the issue and no statements are dropped.
I think documenting this behavior should suffice.