Why Lending Protocol Liquidations Are a Subsidy for Bots

Liquidators must act on public price feeds, creating a race between updating collateral value and seizing it. This forces a trade-off between protocol safety, liquidation efficiency, and user fairness.\n- Safety Delay: Oracle updates are delayed (~1-12 blocks) to prevent flash loan attacks.\n- Efficiency Premium: Bots pay massive gas to win races, passing costs to users.\n- Fairness Failure: Users get zero-price liquidations if bots frontrun the oracle drop.

Liquidation rights are auctioned via gas price wars, consolidating profits into specialized MEV searcher/builder cartels. This extracts value that should recapitalize the protocol, creating a hidden tax on borrowing.\n- Concentrated Risk: A handful of entities (e.g., Flashbots, Jito Labs) control most flow.\n- Protocol Drain: $100M+ in annual liquidation profits are extracted from Aave, Compound.\n- Inefficient Pricing: Liquidators over-collateralize positions to guarantee profit, raising costs for all users.

During market crashes, the outsourced model fails catastrophically. Liquidators face inventory risk and network congestion, leading to under-collateralized positions and protocol insolvency.\n- Adverse Selection: Bots only liquidate easy, profitable positions, abandoning risky ones.\n- Cascading Failure: Mass liquidations cause gas price spikes and oracle staleness, creating bad debt.\n- Historical Proof: ~$100M in bad debt created on Aave/Compound during 2021-2022 crashes.

Protocols like MakerDAO's Collateral Auction Module and Euler's Dutch auctions internalize the process. Collateral is sold at a descending price, ensuring fair market value and disincentivizing pure frontrunning.\n- Fair Price Discovery: Market determines price, not gas wars.\n- Protocol Captures Value: Excess auction proceeds go to the protocol, not bots.\n- Reduced Systemic Risk: Eliminates race conditions during oracle updates.

Shifting from permissionless to permissioned-but-decentralized keeper networks (e.g., Chainlink Keepers, KeeperDAO) aligns incentives. Keepers are slashed for misbehavior and earn predictable fees, not volatile MEV.\n- Aligned Incentives: Keepers are rewarded for system health, not just profit.\n- Reduced Centralization: Broader, staked participant set vs. MEV cartels.\n- Predictable Costs: Fixed fee model makes borrowing costs stable and transparent.

Protocols can segment risk and automate safety mechanisms. Aave V3's Isolation Mode and Compound's borrow caps limit contagion. Circuit breakers (e.g., pausing liquidations during extreme volatility) prevent death spirals.\n- Contagion Firewall: Prevents a single asset's collapse from draining the whole protocol.\n- Graceful Degradation: Automated pauses give oracles and markets time to stabilize.\n- Simplified Complexity: Makes the liquidation surface manageable and auditable.

Traditional first-come-first-served liquidations on Aave V2/V3 and Compound reward pure network speed, not capital efficiency. This creates a zero-sum race where bots invest millions in infrastructure (e.g., Flashbots bundles, mev-geth) to shave milliseconds, extracting value without improving protocol health. The winning bot pays the protocol a fixed discount, pocketing the spread.

Shift from latency races to capital efficiency via sealed-batch or decaying-price auctions. UniswapX and CowSwap prove the model: collect orders over a period (e.g., 5-10 seconds), clear them in a single batch at a uniform clearing price. This forces bots to compete on bid price, not speed, capturing more value for the protocol and liquidated users. MakerDAO's Collateral Auction Module uses a Dutch auction model.

The current model externalizes the cost of bad debt onto the protocol's users while internalizing profits to bot operators. It creates a perverse incentive: bots profit most during high volatility and network congestion, precisely when the protocol is most stressed. This is a direct subsidy from token holders to sophisticated MEV searchers, funded by the liquidation penalty (often 8-15%) paid by users.

Entities like KeeperDAO (now ROOK) and the Flashbots ecosystem are not the root cause but a symptom of the broken mechanism. They emerged to coordinate and privatize this subsidy, demonstrating that the extractable value is so predictable it can be turned into a financial product. Their existence validates the need for mechanism redesign at the protocol layer.

Long-term solutions require architectural shifts. Pre-commitment schemes (like those researched for MEV smoothing) allow liquidators to post bonds and commit to future actions. Combined with sub-second finality from L2s like Solana or Monad, the latency advantage collapses. The goal is to make the liquidation market thick and continuous, not a discrete, winner-take-all event.

Redesigns introduce new risks: oracle manipulation in batch periods, auction griefing, and smart contract complexity. However, the status quo's cost is a persistent leak of protocol value and centralization risk in liquidation markets. The trade-off is clear: accept higher engineering complexity to recapture the $100M+ annual subsidy and build a more resilient, fair system.