EVM WebSocket Gap Recovery: Never Lose Blocks or Logs After Reconnect

Subscriptions are low-latency notifications, not a durable event ledger.

Model the failure correctly

Checkpoint the last fully committed block number and hash, not merely the last received message.

Collect evidence

On reconnect, compare the checkpoint hash with canonical state and roll back to a common ancestor if necessary.

Build the control path

Capture the new head, backfill bounded eth_getLogs ranges, then drain buffered live notifications.

Handle edge cases

Deduplicate by chain ID, block hash, transaction hash, and log index.

Recovery procedure

Adapt range width on provider limits and use jittered retries without multiplying total load.

Production metrics

Test disconnects, duplicate and out-of-order events, failover, busy blocks, and shallow reorgs.

TierZero can help implement or review this architecture through our relevant engineering service.

Primary reference

See the official technical documentation and pin the exact API behavior used by your deployment.

Discuss your system with TierZero.