Skip to main content
Replay checks verify that your node returns correct results. In simple terms: we compare your node’s response to what we observe from the network. If your node consistently disagrees with that baseline, it may be marked as having failed a replay check. We run these checks on read-only, deterministic RPC methods. State-changing or non-deterministic methods (e.g. sends, subscriptions, filters) are not replayed, since their results are not comparable across nodes in the same way.

​
How they work (at a high level)

  • These checks run asynchronously and do not add latency to live consumer requests.
  • Failures are counted per node (and per batch/time window in the portal). Repeated failures can lead to the node being temporarily taken out of production.
  • Common causes: misconfigured nodes, sync issues, or bugs in how certain methods or parameters are handled.

​
What you can do

Ensure your RPC endpoints return correct, deterministic results for the methods and chains you support. If you see replay check failures in Inspect, investigate that node and fix configuration or sync issues. See When checks fail for what happens after a failure and how to monitor recovery.