That is not necessarily true. If the network propagation delay and validation time can be guaranteed to be fast enough, the block time can be lowered to what ever value the values allow for.

The issue is that full nodes would have to commit to having certain levels of connectivity and hardware. Essentially peers would need to talk directly (fully connected mesh network topology) to each other and so peers with low connectivity would suffer, but not the other peers. Thus the number of full nodes would need to be limited.

My insight is that the fully connected mesh connections between pools can be very high quality, while the connections between pools and their miners can be of varying quality. We should not look at pools as bad, rather we only need to limit their sizes so they don't overly centralized the network. Balance.

Payment processors also have to commit to certain high standards of connectivity and hardware.