If a MySQL is producing events faster than connector/Kafka is able to consume it would be useful to provide information about increasing number of unconsumed events (effectively replica lag) so a human operator can react to the situation before binlog segments are deleted by the database.