Leios benchmarking shifts focus from transaction count to transaction size

What happened

A Leios devnet benchmarking harness now scales byte throughput by increasing the number of outputs in each transaction rather than increasing the number of transactions. The change follows a transaction-per-second ceiling set by maxTxsPerEb: each transaction occupies the same 36-byte hash-and-size pair in an endorser-block body regardless of the transaction’s own size.

The record says byte throughput rises linearly as transactions become larger, until the 12 MB maximum endorser-block closure size becomes the limiting boundary. It describes roughly 864 bytes per transaction at that limit, near the stated mainnet median.

Why it matters

This reframes the relevant benchmark from transaction count to transaction bytes. For builders and protocol observers, a fixed transaction-count ceiling can leave room to test more byte throughput through larger transactions, rather than more transactions.

The change is a measurement and workload-generation adjustment, not evidence of a released protocol change or mainnet capacity increase. It is drawn from a thin, single-source set of dev-lane commits.

What to watch

The next useful receipt is benchmark output showing where validation-time capacity binds as output counts rise, and how that compares with the 12 MB closure-size limit. That would clarify whether the byte-size boundary can be reached under the tested workload.