Vitalik Buterin has discussed the potential of 'Keyed Nonces' in enhancing protocol-level support for on-chain privacy solutions and as a significant direction for Ethereum's future state scalability. According to Foresight News, this approach involves creating specialized storage types for specific use cases, aiming to achieve extreme scalability while maintaining protocol decentralization.
For instance, in privacy transactions, nullifiers are used. If on-chain privacy transactions maintain 2000 TPS over eight years, approximately 500 billion nullifiers would be generated. Since nullifiers cannot be pruned, placing them all in the existing general state would exert immense pressure on node storage and decentralization. A dedicated nullifier storage could leverage its highly restricted usage (only for validity checks, with transactions needing to explicitly provide nullifier IDs) to achieve a more decentralized processing solution, such as sharded storage (where nodes only need to hold a small number of shards and maintain connections with honest nodes in each shard) and Bloom Filter optimization (which can significantly reduce verification overhead to about 8 bits per nullifier).
Continuing to rely on the current state may require builders to download tens of terabytes of full data, and privacy users would be unable to efficiently use mechanisms like FOCIL. Therefore, most use cases should migrate to more specialized, cost-effective storage types, while retaining a complete dynamic state for applications like DeFi that require high flexibility, thus achieving a better balance between scalability and decentralization.