Avalanche validators poll a random handful of peers instead of the whole network Classical consensus has every validator hear from every other, so message counts explode as the network grows. Avalanche (@avax) runs the poll differently: each validator asks a small random sample of peers for their preference, switches its own if the sample leans the other way, then repeats. Round after round, the network tips toward one answer. Per Avalanche's docs, the messages a node sends per query stay constant whether the network has twenty validators or two thousand. The protocol also rests when there is nothing to decide, and Snowman, its linear-chain version, is what runs all three chains of avalanche-2:native's primary network.