Liquidations in decentralized lending protocols are often described in deceptively simple terms: monitor health factors, execute the liquidation call, capture the spread. But this framing glosses over nearly everything that separates theoretical understanding from operational reality. A detailed technical breakdown of running a liquidation system across Aave V3 and V4 reveals the sophisticated infrastructure required to compete effectively in this market, touching everything from state computation to MEV routing strategies.
The foundation of any liquidation operation is real-time position tracking. Rather than querying chain state repeatedly through RPC calls—a bottleneck that would limit throughput to thousands of evaluations per block—the most efficient systems compute health factors entirely offline using locally cached data: scaled balances, reserve indices, and current prices. When a price update arrives, typically via a Chainlink oracle, the system evaluates roughly 19,000 positions in milliseconds through pure arithmetic. This architectural choice fundamentally changes what's possible. The challenge intensifies when decoding oracle updates themselves, particularly Chainlink Smart Value Recapture frames, where the data structure requires parsing as function parameters rather than as a simple tuple, and filtering by selector alone misses critical variations in message composition.
Execution pathways diverge sharply across networks. Arbitrum and Base enable Atlas solver operations, introducing constraints around signer relationships, gas ceilings, and the operational requirement for two bonded wallets when managing parallel auctions. Ethereum demands a different approach entirely, leveraging MEV-Share through hash-referenced bundles and builder multiplexing—not as an optional optimization but as a necessity. Arbitrum's Kairos provides Timeboost with 60-second rounds and an express lane, fundamentally altering the latency calculus. Direct execution paths exist where no auction mechanism applies, and their gas profiles reveal behavioral information about which actors rely on these channels.
The economic layer ties these mechanisms together. Liquidation bids must account for precisely where protocol fees are discounted versus where they must remain fixed, and a coupling between two bid parameters creates a silent size floor for the entire system. On-chain execution itself introduces complexity: flash loan cascades that enable leveraged liquidations, provider fee collection mechanisms that operate through incompatible standards, and explicit swap routing that requires careful optimization. These technical and economic dimensions interact in ways that only become visible through months of operational experience. As protocols evolve and MEV mechanisms mature, the sophistication required to compete in liquidations continues rising, making transparent documentation of these systems invaluable for the ecosystem.