On discharge. I was building a complete system for a client. I was testing my "worst case scenario" to make sure that the bank could output at inverter full load until empty. This was an 8 battery 48v system, so based on the battery ratings it should have been capable of charging @100a and discharging @200a. O was testing the system with a 100a load on discharge. I tested it 3 times the first 2 w/o an active balancer and the third time with cell level active balancers inside every battery. I trickle charged the individual cells until the where nearly perfectly balanced, the internal bms was not cutting out until the whole battery voltage was 14.4 or higher. I then slow cycled the whole bank twice, to make sure they where staying balanced. I then torture tested the system with a 100a continuous load a third time. Every time I tested a bms overheated and blew a fet. Soc was between 30% and 50% when the fets blew. I contacted the seller and we reached what I considered an equitable arrangement for a partial refund so I could change the bms's to something that worked better. He did offer a full refund on my entire pallet load, which I declined mainly because I had a deadline to meet vis a vi system install date. I realize my testing was extreme, however it was only @50% of the continuous rating and I felt necessary before I installed a system that I was warrantying since the system was capable of being used in this manner and the client has teenage children. (only a matter of time until a couple curling irons get left on all night, or ...) conversely I tested multiple individual cells @87a discharge full cycle and they performed fine. I finally removed the bms's from the individual cells and added external bulkhead aircraft connectors to each cell to enable me to hook the entire bank into 1 external bms. I used a somewhat rare 6s-20s 200a jbd bms. I then performed the same test 7 more times and it passed every time.