I've seen downstream services fail and RMQ have to swap hundreds of millions of messages to disk, and then resume by itself as consumers picked up... maybe this

I've seen downstream services fail and RMQ have to swap hundreds of millions of messages to disk, and then resume by itself as consumers picked up... maybe this issue you see affects certain versions?