Customers already implementing use cases that involve sending NON_PERSISTENT EMS messages to an FTL infrastructure set up with a peer-to-peer protocol where messages can be sent in bursts will observe a difference in behavior starting with EMS 8.6.0. In that situation, such bursts could increase the EMS server memory utilization. If these bursts are susceptible to exhaust the memory available, this can be addressed in two different ways: 1. If the EMS messages are NON_PERSISTENT because losing them is allowed, you can control memory utilization by setting a value for either the maxbytes or the maxmsgs property of the exporting topic along with setting its overflowPolicy property to discardOld. 2. If the EMS messages should not be lost then they should be sent with the PERSISTENT delivery mode. In that case, you can control memory utilization by enabling flow_control in the EMS server and setting a value for the flowControl property of the exporting topic.