True, but it's O(n^2) complexity instead of O(n) complexity (1000 shards would lead to 1M reads instead of 4k)

If they're not persisted at the receiving side, how does it handle a crash before committing on the receiving node ? Keeping previous versions indefinitely on sending nodes to permit requesting the old values doesn't work, so this introduces a time bound on how long a crashed node has to recover (granted, this could still mean weeks)