Blockchain scalability hierarchy: Computation > Data > State Computation is easier to scale than data. You can parallelize computation, request various "hints" from block builders, or directly replace any amount of computational data with proofs. Data lies in between. If data availability is required, then it must be guaranteed; there is no other way. However, you can split it and use erasure coding, like in PeerDAS. You can achieve graceful degradation: if a node's data capacity is only one-tenth that of other nodes, it can always generate blocks one-tenth the size of the other nodes. State is the most difficult to scale. To ensure that even a single transaction can be verified, you need the complete state. If you replace the state with a tree and retain the root node, then you need the complete state to update the root node. While there are ways to split the state, this involves architectural changes and is not a universal approach. Therefore, if it is possible to replace state with data (without introducing a new form of centralization), this approach should be seriously considered by default. Similarly, if it is possible to replace data with computation (without introducing a new form of centralization), this approach should also be seriously considered by default.