If you're indexing a blockchain, why would you not just read the node database directly via shared memory? @runninyeti.eth would love your thoughts. Reth docs strongly advocate against this.