Optimism/Base unsafe L2 blocks reorg with no sequencer action — every node's derivation pipeline prefers L1 batch data

Symptom for indexers and bots on OP Stack chains: blocks and logs you already ingested at the chain head disappear and are replaced. There is no announcement, no fork, and the sequencer did nothing special.

Cause: every op-node (sequencer included) derives the *safe* chain from batches posted to L1. Unsafe blocks published by the sequencer sit on top of the safe head and are provisional. If the payload attributes derived from L1 batch data fail to consolidate with an existing unsafe block, the derivation pipeline keeps the L1-derived version and reorgs the unsafe blocks away. Because the sequencer runs the same pipeline, it also reorgs and then rebuilds unsafe blocks on the new safe head. No node has to notify anyone — this is just the fork choice rule doing its job, quite unlike a coordinated L1 fork.

Practical consequences: - Treat unsafe-tier data as provisional. If you need stability, read with the block tags safe (L1-derived) or finalized (L1-finalized), accepting the latency. - Make ingestion idempotent and reorg-aware: key rows by block hash, verify parent-hash continuity on each new block, and support rolling back a range when continuity breaks. - Do not rely on transaction index or receipt values from unsafe blocks for irreversible actions such as crediting user balances.

Applies to OP Stack chains (Optimism, Base and derivatives) per the derivation spec as of 2024-2025; the tiering model has been stable, but check the current spec for changes to consolidation rules.

Source: adapted from a Stack Exchange answer by johnhckuo (question by johnhckuo), https://ethereum.stackexchange.com/questions/164412/reorg-on-optimism-base — CC BY-SA 4.0