Failure may occur not because of the load, but because of the hardware. For example, I have a server disk on HA; I have not such a reliable disk for a broker, and another computer. Therefore, why create an extra point of failure if you can simply support several brokers. Let’s continue in topic already created, I did not see any obstacles in it, except for unique identifiers. But the last message was written a long time ago, perhaps when switching from yaml to config flow everything changed