Liquidation Queue

Positions are liquidated in the order they become eligible, based on their health factor or collateral ratio falling below a predefined threshold. This deterministic ordering ensures fairness and predictability, preventing a 'race to the bottom' where liquidators compete for the most profitable positions first.

• Example: If Position A's health factor drops below 1.0 at block 100 and Position B's drops at block 101, Position A will be offered to the queue first.

The queue aggregates multiple undercollateralized positions and processes them in discrete batches at regular intervals (e.g., per block or per hour). This batches sell pressure, preventing a sudden flood of collateral onto the market that could cause slippage and cascading liquidations. It allows the market to absorb sales more efficiently, protecting both the protocol and remaining users.

The queue is filled by keepers or liquidators who purchase the discounted collateral. A fixed liquidation penalty or discount (e.g., 10%) is applied to the collateral's value, creating the economic incentive. The queue manages access to this incentive, often distributing it to the first liquidator who commits capital to clear the next position in line.

Some queues incorporate auction mechanisms to optimize price discovery. Common types include:

• Fixed Discount: Collateral is sold at a predetermined discount.
• Dutch Auction: The discount starts high and decreases over time until a buyer is found.
• Reverse Auction: Multiple liquidators bid, with the one accepting the smallest discount winning the right to liquidate. The queue manages the timing and sequencing of these auctions.

The queue's behavior is governed by protocol parameters set by governance, including:

• Liquidation Threshold: The health factor level that triggers queue entry.
• Liquidation Penalty: The discount applied to the collateral.
• Batch Size & Frequency: How many positions are processed and how often.
• Maximum Collateral per Batch: Limits to protect market stability. These parameters balance insolvency risk against market impact.

A queue differs from a liquidation engine that uses keeper bots in a free-for-all model. Key distinctions:

• Queue: Ordered, batched, less urgent, aims for market stability.
• Engine: Immediate, competitive, prioritizes speed to minimize bad debt. Protocols like MakerDAO (with its Collateral Auction System) use queue-like mechanisms, while many lending protocols use instant engine models.

Liquidation queues differ across protocols based on core parameters:

• Auction Type: Dutch, English, batch, or fixed-price sales.
• Queue Discipline: FCFS, priority-based (worst health first), or time-delayed.
• Incentive Structure: Fixed discount, decaying bonus, or shared rewards.
• Execution: Permissionless (any bot) vs. permissioned (designated keepers). These choices trade off between liquidation efficiency, market stability, decentralization, and resistance to MEV.

Instead of instantly liquidating a position via a public auction or keeper bot, the protocol places the unhealthy position into a first-in, first-out (FIFO) queue. Liquidators or automated systems can then process these queued positions sequentially. This creates a time buffer that can reduce panic selling and allow for partial, controlled liquidations.

• Key Components: Queue position, processing window, liquidation discount.
• Contrasts with atomic, flash-loan-enabled liquidations common in Aave or Compound v2.

The queue mitigates liquidation cascades and bad debt by smoothing out sell pressure. In volatile markets, atomic liquidations can trigger a feedback loop where one large liquidation drives the collateral price down, causing more positions to become undercollateralized. The queue introduces friction and time, allowing the market to absorb sales and giving borrowers a final chance to add collateral or repay debt before their position reaches the front of the queue.

To ensure queued positions are processed, protocols offer a liquidation bonus or discount. The first liquidator to claim a position from the front of the queue can purchase the collateral at a discount to the market price, pocketing the difference as profit. The size of the discount and the queue's processing speed create a competitive landscape for liquidators, balancing efficiency with system stability. Protocols like MakerDAO (with its Auction Surplus Buffer) and Euler Finance (pre-hack) have implemented variations of this model.

While reducing systemic risk, queues introduce new attack vectors and considerations:

• Queue Stalling: A malicious actor could spam the queue with small, unprofitable positions to delay the liquidation of a large, connected position.
• Oracle Risk: The extended time window increases reliance on price oracle accuracy; a stale price could lead to mispriced liquidations.
• Guaranteed Liquidity: The protocol must ensure there is always sufficient capital (e.g., from a surplus buffer or designated liquidity pool) to cover the discounted purchase, or risk accumulating bad debt.

MakerDAO's Collateral Auction System: Uses a reverse Dutch auction (falling price) with a time delay, effectively creating a queue-like process for vault liquidations. The system includes a surplus buffer to cover any remaining bad debt.

Compound III's Base-Liquidation Model: While not a strict queue, it uses a base asset (e.g., USDC) for all liquidations and a minimum borrow rule, which reduces atomic sell pressure on diverse collateral assets, achieving a similar stabilizing effect.

• Health Factor / Collateral Ratio: The metric that determines when a position is eligible for the queue.
• Liquidation Threshold: The specific collateral value ratio that triggers queue entry.
• Auction Mechanisms: Dutch, English, and sealed-bid auctions as alternative liquidation methods.
• Bad Debt: The ultimate risk a queue aims to prevent, where debt exceeds the value of liquidated collateral.
• Keepers: Automated bots that monitor and interact with the liquidation queue.