We need datapoints to see how the network would degrade when the blocks are consistently way over 1MB. If the impact is minor, then who knows some level of safe block size increase could also be incorporated into core.

I am fine with the idea of using BCH as a guinea pig for big block stress testing.