How exactly did MakerDAO's infamous incident — $8.32 million in collateral selling for 0 DAI — happen?
This incident occurred on March 12–13, 2020, during the market crash commonly known as "Black Thursday." ETH's price fell sharply in a short window, triggering a wave of vaults into Liquidation. MakerDAO's English auction mechanism at the time required liquidators to actively submit a bid transaction and pay gas to participate. But over those two days, the Ethereum network became severely congested from the surge in transaction volume, and gas fees spiked so high that the vast majority of would-be liquidators simply couldn't get their bid transactions included in a Block in time.
In the end, only a single liquidator managed to successfully submit a bid. Under the English auction's "highest bidder wins" rule, even though that bid was as low as 0 DAI, since there was no competing bid, the auction settled according to the rules anyway. MakerDAO ended up losing $8.32 million in collateral outright, recovering none of the corresponding debt. This event directly exposed the structural weakness of English auctions — that they need enough competing bidders to function well — and was the direct catalyst behind MakerDAO launching Liquidations 2.0 and switching to Dutch auctions later that same year.
Why does combining Dutch auctions with flash loans let liquidators participate with zero capital cost?
The key is that a Dutch auction settles instantly — the moment a bid succeeds, collateral and payment change hands right there, unlike an English auction, which requires locking up capital, waiting to see if someone else bids higher, and only confirming the outcome once the auction closes. Because settlement is instant, a liquidator can execute the entire sequence within a single transaction: borrow a large sum of DAI via a Flash Loan (which requires no collateral, only that the principal plus interest gets repaid before the same transaction ends), use that borrowed DAI to buy the discounted collateral in the auction, immediately sell that collateral on another DEX to convert it back to DAI, repay the flash loan principal and interest, and pocket the spread in between as profit.
This means the liquidator themselves never needs to have any starting capital ready — they only need the ability to construct the transaction and pay gas to participate, which strips away the barrier of "only those with capital can liquidate" entirely, turning all the idle liquidity sitting around DeFi into a potential Liquidation funding pool. That's also why pairing Dutch auctions with flash loans made liquidation competition far more intense than it was in the English-auction era — the barrier to entry dropped to almost nothing but technical Skill and gas fees.
Why does Aave cap a liquidator at repaying a maximum of 50% of the borrowed amount in a single Liquidation?
This mechanism is called partial liquidation, and its purpose is to prevent a liquidator from wiping out an entire position in one shot. Without this cap, the moment the Health Factor drops below the threshold — even by a tiny Margin — a liquidator could have an incentive to liquidate the entire position at once and capture the maximum Liquidation Bonus available, which would be excessive punishment for the borrower. You might trip the threshold from nothing more than a brief market wobble, with plenty of account equity still remaining, only to see your position wiped down to almost nothing in a single liquidation event.
Capping a single liquidation at 50% gives the borrower a chance, after triggering liquidation, to catch their breath with the remaining half of the position — to top up collateral or wait and see whether the market stabilizes — rather than being fully closed out instantly. It also creates a more gradual form of competition among liquidators: if the health factor still hasn't returned to a safe range after the first partial liquidation, a new liquidation opportunity opens up, letting the next liquidator step in rather than letting the first liquidator capture all the benefit in one go.
Before choosing which lending protocol to deposit assets into, how should I use "Liquidation mechanism design" as a lens for judging it?
The first step is figuring out whether the protocol uses an auction model or a fixed-discount model — this information is usually in the protocol's developer documentation, searchable under terms like "liquidation mechanism," "auction," or "Liquidation Bonus." If it's a fixed-discount model, look up the actual discount rate number — the higher it is, the larger the proportional loss you'll take if you get liquidated. If it's an auction model (Dutch or reverse Dutch), pay attention to whether the price curve is linear, step-wise, or exponential, since that affects how long a liquidator is willing to wait before stepping in, which in turn affects how much risk your position carries during the auction window.
The second step is checking whether the protocol has actually been tested by real extreme market conditions. MakerDAO's Black Thursday, and multi-hundred-million-dollar liquidation events at Aave and Compound over the years, tell you how the mechanism actually performs under pressure, beyond its design on paper. A liquidation mechanism that's never faced a genuine stress test is still theoretical no matter how thorough the documentation reads — which is why looking up "when was this protocol's last large-scale liquidation event and how did it play out" is worth more than reading the mechanism description in a Whitepaper alone.
Your health factor drops below 1 and your position gets liquidated — but the word "Liquidation" hides an entire auction mechanism running underneath it, and the specific auction design a protocol chooses directly determines how much you lose this time, and who profits from the gap. DeFi lending protocols mainly handle liquidations through two models: Dutch auctions, where the price starts high and falls until someone takes it, and fixed-discount models, where a pre-set discount rate lets liquidators buy collateral outright with no auction process at all. Each design solves a different problem, and each creates a different kind of risk.
MakerDAO originally used an English auction — bids called out from low to high, highest bidder wins. This design had several fatal flaws. The full auction typically took 6 hours, and every participant's capital stayed locked until they were outbid or the auction closed, which discouraged enough people from participating in the first place. Worse, one auction for $8.32 million worth of collateral ultimately settled for 0 DAI, because during the "Black Thursday" crash in March 2020, Ethereum network congestion meant only a single liquidator managed to get a bid transaction through — effectively a monopoly bid. MakerDAO responded by launching Liquidations 2.0 in 2020, switching entirely to Dutch auctions: the starting price is set slightly above the Oracle price, and it continues falling over time until someone is willing to buy at the current price or the debt is fully repaid. Dutch auctions settle instantly, with no capital lock-up waiting for a bidding outcome, which lets anyone participate at zero capital cost via flash loans — the entire borrow, buy-collateral, and repay sequence can complete within a single transaction.
Aave takes an entirely different route — no auction at all, just a fixed liquidation bonus that lets liquidators buy collateral outright at a discount to market price. A liquidator repays some or all of a borrower's debt (Aave caps this at 50% of the borrowed amount per liquidation) and immediately receives an equivalent value of collateral at that fixed discount — there's no price-moving-over-time auction stage at all; the trade settles the instant it's submitted. The advantage is speed and simplicity — liquidators don't need to guess where an auction might settle. The tradeoff is that the discount rate is hardcoded and doesn't automatically adjust to how deep the actual market liquidity is at that moment. When liquidity is good, the fixed discount can end up higher than what a liquidator actually needed as compensation; when liquidity is thin and sell pressure is heavy, the same fixed discount might not be enough to attract a liquidator to step in immediately.
Euler v2 retained the reverse Dutch auction mechanism inherited from v1 as its default liquidation method, while also letting advanced vault creators customize their own liquidation flows. Euler Labs CEO Michael Bentley told The Block that this mechanism is popular precisely because it can offer some of the lowest liquidation bonuses in DeFi, which helps protect borrowers and maintain pool solvency. The logic works like this: the discount starts at zero and gradually rises over time until a liquidator is willing to take the position at the current discount rate. This is the inverse of MakerDAO's "price falling from high to low" — Euler instead uses "discount rising from low to high" — but both are fundamentally the same underlying logic of letting the market discover a price just attractive enough to bring in a buyer. The difference is only in what's being anchored: a discount rate versus an absolute price.
Lining up all three designs side by side reveals a clear spectrum. The fixed-discount model (Aave) sacrifices price-discovery precision for speed and simplicity. The classic Dutch auction (MakerDAO) uses a price that falls over time to approximate what the market would actually accept, but requires time for the price to reach that point. The reverse Dutch auction (Euler) hands off the question of "how low can the discount go while still protecting borrower equity" to market dynamics as well, rather than having protocol governance fix it in a single decision. Academic research confirms this tradeoff exists: an empirical study of DeFi liquidation mechanisms out of UC Berkeley found that fixed-discount models skew notably more toward liquidators than auction-based models do, at the borrower's expense. That means which lending protocol you deposit assets into — and therefore which liquidation mechanism you're subject to — already determines the answer to "how much less will I get back if I'm liquidated" before you ever hit that threshold.