# PayBench calibration provenance — R2 x402-Stellar settlement-finality
# Updated 2026-06-04 with first-party n=30 (MANUAL direct-payment path; spec/facilitator run pending OZ key).

rail: "R2 x402-Stellar"
dimension: settlement-finality
unit: seconds

distribution:
  family: lognormal
  parameters: {median_s: 2.74, sigma_log: 0.116}   # EMPIRICAL n=30, manual direct-payment (no facilitator)
  approach: first-party-empirical

sources:
  - type: first-party
    ref: "poc/rail-stellar-x402 measure-finality-manual.ts, n=30 testnet; samples/R2-x402-stellar.samples.jsonl"
    date: "2026-06-04"
    sample_size: 30
    note: >
      0 failures. Manual direct classic-USDC payment buyer->seller via Horizon submitTransaction
      (resolves at ledger close = finality). median 2.74s, sigma_log 0.116, p50 2.73, p90 3.17, p99 3.48.
      MANUAL path: no x402 facilitator, no Soroban auth-entry — omits the facilitator round-trip.
  - type: documentation
    ref: "Stellar ledger close ~5.8s (Q3 2025 report / dashboard)"
    date: "2026-06-04"
    sample_size: 0
    note: "Deterministic finality at ledger close, no reorg. submit->inclusion averages ~half a ledger interval -> ~2.7s observed."

calibration_date: "2026-06-04"
confidence: high
fixture_content_hash: "sha256:b002e83280933bfebd0aa2b35e3b8610b489c6130c55e5a62f99c095d2a0b38c"
notes: >
  First-party n=30 supersedes the doc seed (6.0 / 0.26), which over-estimated the median (it assumed a
  full ledger close + facilitator; real submit->inclusion averages ~2.7s). This is the MANUAL
  direct-payment path -> a LOWER BOUND on spec x402-on-Stellar finality, which adds the OZ facilitator
  round-trip (verify+settle API + facilitator rebuild/submit). Spec path is wired (server.ts via
  @x402/stellar + OZ facilitator, createAuthHeaders) and pending the free OZ testnet key
  (https://channels.openzeppelin.com/testnet/gen). Re-measure on the spec path once keyed; expect the
  median to rise by the facilitator round-trip.