solana-tx-bench reports

Static reports published from reviewed benchmark artifacts. Provider ACK latency is diagnostic only; landing evidence uses the first matched stream observation by transaction signature. Processed metadata is used for block position when available.

• DoubleZero TPU vs public Internet, paired 1h run, 2026-07-13
  Same-leader causal A/B from Frankfurt across 246 verified DoubleZero-routed leader runs. Includes same-slot PoH placement and a city-level analysis showing where the result is concentrated.
• RPCEdge always-race 1h leader-paced run, 2026-07-03
  Standalone artifact report generated from the Frankfurt benchmark directory only. Includes tail events and unobserved signatures. ClickHouse is not used for this public report.
• RPCEdge paired route-policy 5m smoke with 3x/50ms resend, 2026-07-03
  Same signed caller transaction was submitted at 0ms, +50ms, and +100ms for each policy arm. Landing metrics are deduplicated by transaction signature; provider ACK rows remain attempt-level diagnostics. Includes policy comparison, no-spam vs 3x comparison, tail events, and unobserved signatures.
• RPCEdge cost-aware paired route-policy 5m smoke, 2026-07-02
  Includes policy comparison, tail events, and unobserved signatures. Harmonic-selected transactions were signed with the minimum 300,000 micro-lamports/CU required by Harmonic. This is a smoke report, not the final 1h authoritative benchmark.
• Legacy exploratory paired route-policy 5m run, 2026-07-01
  Kept for tail/debug reference only. Superseded by the 2026-07-02 cost-aware smoke.
• Legacy RPCEdge QUIC zero-priority 1h leader-paced run, 2026-06-30