Liveness Failures in Ethereum Consensus

Validators are economically incentivized to reorder or exclude transactions to capture MEV, directly threatening chain liveness for ordinary users.

• PBS (Proposer-Builder Separation) is a partial mitigation, but builders can still censor.
• Real-world example: OFAC-sanctioned addresses were censored by ~45% of post-merge blocks at its peak.
• This creates a two-tiered system where liveness is a paid privilege.

A critical bug in a consensus client with >33% network share can cause a catastrophic chain split or halt.

• The Geth client historically commanded ~85% dominance, creating a systemic risk.
• A liveness failure here would require a coordinated, manual intervention from client teams.
• This is a failure of decentralization at the infrastructure layer.

Ethereum's last-resort liveness mechanism. If >1/3 of validators go offline, the protocol slashes their stake over days to re-achieve finality.

• This is a cryptoeconomic solution to a coordination problem.
• Ultimately relies on social consensus (Layer 0) for recovery if the leak fails.
• Highlights that blockchain security is a human game with automated tools.

A 67% adversarial supermajority can theoretically revert finalized blocks, destroying the core guarantee of settlement.

• This is considered a catastrophic liveness and safety failure.
• The cost is astronomically high (~$20B+ in slashed ETH), making it a near-theoretical attack.
• The real threat is a bug or exploit that lowers this cost.

Mitigates supermajority client risk by enforcing a <33% max share for any single execution or consensus client.

• Post-merge efforts have successfully reduced Geth's share to ~65% (still critical).
• Requires ongoing grant funding and validator education to maintain balance.
• Tools like Ethereum Client Diversity dashboard provide critical visibility.

Lido, Coinbase, Binance control ~45% of all staked ETH. While technically decentralized, this creates a latent liveness threat.

• Coordinated action (or regulatory pressure) on a few large entities could threaten the >33% inactivity threshold.
• This represents a political and regulatory attack vector orthogonal to code.
• The solution is not technical, but economic and governance-based.

Ethereum's Gasper consensus provides probabilistic finality. A block is only considered finalized after ~12.8 minutes (2 epochs). Before that, reorgs are possible. This creates a race condition for DeFi arbitrage, bridge attestations, and NFT marketplaces that assume immediate settlement.

PBS externalizes block production to a competitive builder market via MEV-Boost. This introduces a new liveness vector: the proposer (validator) must be online and connected to relays to receive the best block. A validator going offline defaults to local, less profitable block building, degrading chain quality.

Validators can censor transactions by excluding them from blocks. While protocol-level enforcement (e.g., proposer commitments) is weak, the real threat is OFAC-compliant relays like Flashbots, which filter sanctioned addresses. This creates liveness failures for specific users/applications, breaking atomic composability.

The long-term fix is enshrined PBS (eProtocol) moving the builder market into the core protocol. Short-term, crLists (censorship resistance lists) allow users to force transaction inclusion. Architects must design for inclusion guarantees, not just execution speed, integrating with services like Flashbots Protect.

Correlated client bugs (e.g., Prysm, Teku) or cloud provider outages (AWS) can cause mass slashing or inactivity leaks, threatening chain liveness. This systemic risk is why client diversity is a security parameter, not just a nice-to-have. A >33% client share is a single point of failure.

Protocols must monitor client diversity metrics and be prepared to operate through a liveness failure. This means designing for extended finality delays and having contingency plans for using alternative data layers (e.g., EigenLayer, Lagrange) if the primary chain halts. Don't assume 24/7/365 liveness.