Skip to main content

Reliability Model

  • Audience: Trading API integrators.
  • What this page covers: What API users can expect about execution responses, data freshness, retries, and settlement visibility.

Reliability Mental Model​

For API users, reliability is best understood as three separate timelines:

  • execution timeline: when an order is accepted, rejected, filled, or cancelled
  • query timeline: when that result appears on order-history and market-data endpoints
  • settlement timeline: when the rollup publishes state commitments and proofs to VSL

These timelines are related, but they are not always simultaneous.

Guarantees vs Non-Guarantees​

Generally expected​

  • A trading response reflects the execution decision for that request.
  • OMS and market-data endpoints converge to the same trading outcome after short propagation delays.
  • Rollup state is published to VSL on a recurring cadence of every 2 blocks.

Not guaranteed as immediate​

  • Zero-latency synchronization across all public endpoints.
  • Instant visibility of a new order, fill, or cancel across every query surface at the same moment.
  • Immediate onchain publication for every individual trade event.

Short-lived eventual-consistency windows are normal in distributed async pipelines.

Integration Best Practices​

  • Use client-generated order identifiers or request correlation where your integration supports it.
  • Treat OMS and market-data as eventually consistent views of recent execution activity.
  • After reconnecting, resync open orders, balances, and market state before resuming automated strategies.
  • Build clients to tolerate brief ordering differences between trading responses and downstream market-data updates.

What Finality Means Here​

  • API acceptance or rejection tells you what the execution layer decided for the request.
  • OMS visibility tells you the read model has caught up enough for user-facing queries.
  • VSL publication is the public verification step for the rollup state update.

For integrators, the key distinction is simple: trade execution is fast, public verification follows on the VSL cadence, and read APIs may briefly trail the execution response.

Cross-check behavior in Engine API, OMS API, and Market Data API.