16835

Repeatedly computing the same Merkle root makes tracing the data flow more interesting than adding more threads. Agave Draft #16835 proposes passing along the shred roots already computed during signature verification, together with the verification results, avoiding recalculation downstream and a dedicated thread pool. The original text also explains the behavioral boundaries: only shreds that can be parsed and pass sanitize are forwarded, while the range eligible for block inclusion remains unchanged, since the insertion path already filtered them. In my view, optimization requires tracking duplicated work as well as checking where filtering occurs. This proposal, $SOL , still has node operation and analysis to complete, so it does not justify claiming a mainnet speedup percentage as though it had already been implemented.