When the Latency high mark is achieved, incoming write requests are stalled, till the pending number of updates reaches the preset Latency_low_mark value. This is done to ensure that the high mark is not crossed too frequently. Latency protection is required to make sure that the Secondary does not lag behind too much, in case of heavy Input/Outputs or poor network in the asynchronous mode of replication. For more information refer to Latency_high_mark.