(Alex Poritskiy) Recently I worked on one customer issue that I would describe as “slave performance too slow”. During a quick analysis, I’ve found that the replication slave SQL thread cannot keep up while processing row-based events from the master’s binary log.