linking to the forum post that proves out and measures the stale block and fork rate after reducing the block times.
if we configure tcp correctly, we can maintain an under %5 stale block rate even in the worst case conditions