Why is Walrus's recovery cost "O(B) across the entire network"? @Walrus 🦭/acc 's Red Stuff uses a two-dimensional sharding structure, where each node only stores O(B/n) data. When a node fails, repairing it only requires pulling "shard fragments" from a small number of neighbors, rather than the entire file. More critically: the repair traffic does not grow linearly with the size of the network! This allows Walrus to maintain low bandwidth overhead even with tens of thousands of nodes — truly scalable decentralized storage. $WAL #walrus 🌐