if I run 5 nodes across the world,
and each nodes has millions of transactions
and the latencies between pair of nodes are different,
when multiple transactions are recorded on node1,
can I guarantee the blocks will be distributed well with strong reliability?