It would be a recognition that the block size limit is
not part of the consensus layer, but rather part of the transport layer, as sickpig suggested:
you know what I can't stop thinking that the max block size is a transport layer constraint that have crept in consensus layer.
The network would dynamically determine the max block size as the network evolves by expressing the size of the blocks they will accept with the drop-down menu on their client.
This seems too easy, like why wouldn't this have been thought of before. Is the idea that maybe this is one of those cases where muddled thinking (the consensus/transport layer confusion) has prevented people from seeing the obvious? I ask because I'm not sure I understand the full implications of sickpig's comment.