Thanks, I can reproduce it now. The reason it was not reproducible in the MTR test is that MTR sets up the slave with binary log and log_slave_updates enabled by default. With those options disabled and slave restarts removed, the insert is indeed not blocked by an active BINLOG lock.
I could also reproduce the lock issue with SHOW SLAVE STATUS, but will report it separately.
Thanks, I can reproduce it now. The reason it was not reproducible in the MTR test is that MTR sets up the slave with binary log and log_slave_updates enabled by default. With those options disabled and slave restarts removed, the insert is indeed not blocked by an active BINLOG lock.
I could also reproduce the lock issue with SHOW SLAVE STATUS, but will report it separately.