Freshness by channel
BlockDB processes datasets in a coordinated pipeline, so most tables advance together. The main differences you observe in freshness come from delivery channel and chain-specific reorg/finality buffers.Real-time (WSS)
- Lowest latency delivery.
- Reorg-aware: you may see updates for recently produced blocks within the real-time buffer window.
API (REST)
- Near-real-time, but with stability guarantees.
- The API keeps history and applies a chain-specific backoff beyond the maximum expected reorg window, so results are stable and reproducible when you query the same range again.
Warehouse shares
- Near-real-time to hourly, depending on the provider and replication cadence.
Bulk exports (SFTP/S3/Blob)
- Nightly snapshots by default, with ad-hoc hotfix replays when required.
Chain buffers (blocks)
| Chain ID | Chain | Archive/API buffer (blocks) | Real-time buffer (blocks) | Evidence & rationale | Source |
|---|---|---|---|---|---|
1 | Ethereum Mainnet | 64 | 0 | Ethereum finalizes a block after ~2 epochs (≈64 blocks, ~13-15 minutes). | https://www.circle.com/blog/exploring-confirmation-rules-for-ethereum |
10 | Optimism | 600 | 0 | Optimism finality is reached when the underlying L1 block is finalized (~20 minutes). At ~2s per L2 block, ~600 blocks ≈ 20 minutes. | https://docs.optimism.io/op-stack/transactions/transaction-finality |
56 | BNB Chain | 120 | 0 | BNB fast finality finalizes block n by n+2 (~3.75s). When unavailable, probabilistic finality can require ~120 blocks (~180s). | https://docs.bnbchain.org/bnb-smart-chain/developers/json_rpc/bsc-api-list/ |
130 | Unichain | 1440 | 0 | Chainlink lists Unichain finality at ~24 minutes. Assuming ~1s blocks, ~1440 blocks ≈ 24 minutes. | https://docs.chain.link/ccip/ccip-execution-latency |
137 | Polygon | 500 | 5 | Chainlink specifies Polygon PoS finality at a depth of ~500 blocks (~17 minutes). | https://docs.chain.link/ccip/ccip-execution-latency |
146 | Sonic | 10 | 0 | Sonic targets sub-second deterministic finality with no rollbacks. | https://docs.soniclabs.com/technology/overview |
5000 | Mantle | 840 | 0 | Chainlink reports Mantle finality at ~28 minutes. With ~2s blocks, ~840 blocks ≈ 28 minutes. | https://docs.chain.link/ccip/ccip-execution-latency |
8453 | Base | 600 | 0 | Base finality is ~18 minutes. At ~2s per block, ~600 blocks ≈ 18 minutes for API stability. | https://docs.chain.link/ccip/ccip-execution-latency |
42161 | Arbitrum | 4080 | 2 | Chainlink lists Arbitrum finality at ~17 minutes (~4080 blocks at ~0.25s). Transactions become irrevocable once the underlying L1 block is finalized. | https://docs.chain.link/ccip/ccip-execution-latency |
43114 | Avalanche C-Chain | 3 | 0 | Avalanche provides deterministic finality in ~1-2 seconds (Snowman). | https://www.galaxy.com/insights/research/ready-layer-one-avalanche |
59144 | Linea | 600 | 0 | Linea finality requires ~600 blocks (~20 minutes). | https://docs.chain.link/ccip/ccip-execution-latency |
These values are operational defaults and may change over time as chains evolve. If you need stricter guarantees for a specific chain, contact support.
How do I monitor data freshness?
- Use the
_updated_atcolumn for the dataset you query. - For API pagination, you can also monitor the response
cursorprogression and returnedcountover time.
Need tighter guarantees? Mirror the dataset via the API and compare
_tracing_id values. Differences indicate your mirror is stale or queried inside the reorg buffer.Backfill & replay behavior
- Hotfix windows — If an upstream outage or a downstream bug is found, BlockDB will replay the affected ranges.
- Historical expansions — New datasets start with block 0 and backfill to present.