Why Ethereum Reorgs Happen After the Merge

Proof-of-Stake consensus eliminated the economic incentive for deep reorgs. Under PoW, miners could profitably re-mine blocks. PoS validators are slashed for equivocation, making deep reorgs economically irrational and protocol-breaking.

Network latency creates natural forks. Validators in different geographic regions receive blocks at different times. This causes temporary forks where two competing chains exist. The fork-choice rule (LMD-GHOST) selects the heaviest chain, orphaning the lighter chain.

Single-slot finality is the endgame. Current reorgs occur because Ethereum has probabilistic finality. Proposals like Ethereum's single-slot finality (SSF) will make reorgs after one confirmation impossible, mirroring the guarantees of Solana or Aptos.

Evidence: Post-Merge, 90% of reorgs are 1-2 blocks deep, caused by latency. The Ethereum Beacon Chain averages a 7-block reorg every 9 days, a fundamental latency tax on applications like Uniswap and AAVE that rely on immediate state.

Post-Merge, reorgs are consensus. Under Proof-of-Work, a reorg signaled a failed 51% attack. Now, with Proposer-Builder Separation (PBS), temporary forks are the network's method for resolving honest disagreements between validators. The fork choice rule (LMD-GHOST) actively expects and resolves them.

MEV drives reorg economics. Builders like Flashbots and BloXroute compete to propose the most profitable block. If a builder sees a more lucrative block arrive late, the economic incentive to reorg is inherent. This creates a reorg-for-profit market, not an attack.

Finality is the new security. The old security model relied on probabilistic settlement. The new model uses checkpoint finality. A block is only truly secure after two epochs (~12.8 minutes), when it's cryptographically finalized. Everything before that is subject to reorg.

Evidence: Post-Merge data from EigenPhi and Rated Network shows reorgs of 1-2 blocks are routine, occurring daily. The record is a 7-block reorg, demonstrating the system's designed tolerance before finalization kicks in.

Proposer-Builder Separation (PBS) is the root cause. Validators now outsource block construction to specialized builders via MEV-Boost relays like Flashbots and BloXroute. This auction introduces a latency gap between seeing a block and confirming it.

The 1-block reorg window opens when a builder with a more profitable block arrives late. A validator will orphan the first block for the higher payout. This is a rational, profit-driven action, not a consensus failure.

Reorgs are now economic events. Pre-Merge reorgs required 51% hash power. Post-Merge, a single validator can reorg the prior slot by proposing a more valuable block, a tactic leveraged by tools like Jito's bundle auctions on Solana.

Evidence: Ethereum's reorg rate increased 5x after the Merge. Builders like Titan Builder optimize for this window, creating blocks so profitable that validators willingly reorg to claim them, proving the market's efficiency.

Finality is probabilistic. The Merge replaced PoW's 'longest chain' rule with PoS finality, but it's not instantaneous. Validators vote on checkpoints, and finality is only achieved after two consecutive epochs (~12.8 minutes). This creates a reorg window before blocks are finalized.

Network latency creates forks. Even with honest validators, global network propagation delays cause temporary forks. Tools like MEV-Boost exacerbate this by introducing relay latency, allowing competing blocks to be built on different chain heads before gossip propagation resolves the fork.

The reorg risk is quantifiable. The probability of a reorg decreases exponentially with block depth. A 1-block reorg has a non-zero chance; a reorg of finalized blocks requires a coordinated 1/3+ validator attack, which is economically prohibitive due to slashing penalties.

Infrastructure must account for this. High-value applications like Across Protocol and Chainlink use multi-block confirmations. Builders must design assuming the canonical chain is only probabilistically certain, not a guarantee from the first block.

Reorgs are consensus latency. The Merge eliminated proof-of-work's probabilistic finality, but Ethereum's 32-slot finalization window creates a deterministic attack surface. An attacker with 33% of staked ETH can stall finality, enabling short-range reorgs within this window.

Single-slot finality (SSF) eliminates this. It replaces the Casper FFG finality gadget with an instant, single-slot mechanism. This removes the reorg window entirely, making chain reversion computationally infeasible after a single block.

The trade-off is validator load. SSF requires all validators to sign every block, not just committees. This demands massive improvements in signature aggregation and network gossip, a core focus of Ethereum R&D teams like the Ethereum Foundation and ConsenSys.

Evidence: The Inactivity Leak. This mechanism, triggered after four epochs of non-finalization, demonstrates the cost of delayed finality. It systematically burns the stake of offline validators to restore a finalizing supermajority, a safety net SSF renders obsolete.