ezETH Liquidation Cascade, April 2024

A Liquid Restaking Token that could not yet be redeemed had only one observable price: a pool a few thousand ETH deep. Lending markets priced collateral from that pool, and the liquidators those markets sent in had to sell the seized collateral back into it.

1. Ninety minutes, mostly in twenty-three

Gauntlet, the curator of the Morpho vault that supplied most of the WETH in the affected market, published a timeline the same day. The ezETH/WETH oracle first fell below 1.0 at 02:25 UTC. The first liquidation landed at 02:29:47, four minutes later, in block 19,722,269. By 02:52:35 — twenty-three minutes after the first — 145 of the day's 146 liquidations were done. The last one came at 13:00:47.

Across the two markets, 138 liquidations hit the 0.86 LLTV market and 8 hit the 0.77 market. On the first, 5,035 WETH of debt was repaid and 6,236.72 ezETH of collateral seized, leaving 10.96 WETH of insolvencies socialised to lenders. On the second, 804 WETH repaid, 1,000.25 ezETH seized, and nothing left unpaid. Gauntlet notes that only borrowers in the 86% market accumulated insolvent debt, and puts its own vault's share of the loss at 7.12 WETH, or 11 basis points of that vault's WETH supply.

At Gearbox, leveraged ezETH positions were liquidated too. Its own write-up five days later reports no bad debt for passive lenders and a profit for the protocol, and gives the event's ledger: 90 ETH of liquidation fees, 25 ETH of temporary downside, and 170 ETH of premium earned by the emergency liquidator. The DAO then voted to hand 235 ETH of that back to the liquidated borrowers — while saying in the same breath that this should not become a precedent, because leverage carries risk.

2. The loop

The part worth understanding is why this accelerated instead of settling.

The oracle pricing the Morpho market did not read a single venue. Gauntlet lists four: an ezETH/WETH Balancer pool, an ezETH/weETH/rswETH Balancer pool, an ezETH/WETH Curve pool, and an ezETH/WETH Uniswap V3 pool. Aggregating venues is the standard defence against someone pushing one market around.

It does not help when every venue is draining at once. WETH liquidity in the ezETH/WETH Balancer pool fell from about 4,000 WETH to as low as 500 before recovering to 1,300 over eight hours. Gauntlet attributes the drain to two things: holders swapping out of ezETH, and — this is the important one — liquidators swapping out of the ezETH they had just claimed.

That closes a circuit. Thinner pools pull the oracle down; a lower oracle triggers liquidations; each liquidation hands a liquidator collateral they must sell to realise their profit; they sell it into the same four pools the oracle is reading; the oracle falls further. Gauntlet states the sequence plainly, and notes that liquidity in that Balancer pool remained 65% below pre-event levels even after partially recovering.

Liquidation is usually described as the mechanism that protects lenders from a falling price. Here it was also a mechanism that lowered the price.

3. Why the safer oracle was not available

Both venues reached the same conclusion, and both said so against their own interest.

Gauntlet, asked whether the vault would move to markets using an exchange-rate oracle, answered that it was looking into it and that such an oracle would likely have led to fewer borrowers being liquidated. Gearbox went further in its own post, arguing that oracles reading an asset's redemption value rather than its market price reduce liquidation volatility without introducing hard trust assumptions, and that this applies to liquid restaking tokens — with a conditional attached: the ones that have redemptions.

ezETH did not. Renzo's own announcement at the time put the upgrade enabling ezETH withdrawals in May 2024, and Gearbox's post, published on 29 April, warned its users to hold a health factor of 1.2 or better to survive price swings that could occur "before these protocols enable withdrawals".

So the causation runs in one line. No redemption path meant no redemption value to read. No redemption value meant the only available price was whatever the pools said. And the pools were where the liquidators were selling.

That condition has since changed: Renzo's documentation now describes withdrawals as enabled on Ethereum mainnet, taking up to fifteen days. The structural hazard this case file describes applies to a receipt during the window before redemption exists — which is exactly the window in which new receipts are promoted hardest.

4. A safety mechanism that cost money

One detail deserves its own heading because it is so easy to get backwards. Gearbox had a DAO-approved pause role that triggers on certain conditions, and the cascade tripped it. The pause worked. It also, as Gearbox itself records, disabled every protocol function — so borrowers could not add collateral to rescue their own positions, and some liquidations that need not have happened did. During the pause only the emergency liquidator could act, which is why that liquidator collected the 170 ETH premium the DAO later voted to redistribute.

Gearbox's listed remedies included allowing collateral top-ups during a pause, and partial rather than full liquidation. Neither is an exotic idea. Both were absent on the day.

5. Where the sources disagree

• How far the price fell. Not a narrow spread but an irreconcilable one. The Block reported ezETH losing "up to 18%" of its parity with ETH; Cointelegraph reported a low of $688 on Uniswap, which against an ETH price near $3,100 that week is a fall of roughly three quarters. Both cannot describe the same quantity. The likely reading is that one is a momentary print in a single drained pool and the other a broader measure, but no party published a figure to settle it. Either way it is not the number that caused the liquidations, because the lending markets priced from feeds aggregating four pools, and the only threshold Gauntlet reports is that its oracle fell below 1.0. This page uses the oracle behaviour and leaves the headline price alone.
• The total damage. Circulating aggregates run from more than $50M to over $65M, with per-venue splits that do not reconcile between outlets. No party published a combined figure. The numbers in section 1 are the ones the parties published about their own markets, and they are deliberately not summed here.
• Which oracle moved first. Gauntlet's appendix reports that RedStone and Chainlink feeds were largely in sync but that RedStone responded faster and posted more updates in the window. Gearbox's contemporaneous statement attributed the depeg it saw to the RedStone feed. These are compatible, but note that the venue naming a specific feed was the one using it.
• Why ezETH was being sold. Reporting ties the sell-off to the REZ token distribution. The Block adds a specific mechanism: that Renzo disqualified airdrop farmers who had been looping — selling ezETH for ETH and redepositing to accrue more rewards — and that those users then exited by selling ezETH in quantity. Renzo subsequently raised the first-season allocation from 5% to 7% and moved the claim date to 30 April. If that account is right it is notable, because it would mean the positions unwound were the leveraged ones, and leveraged positions are precisely what the lending markets held. But no party published an account of the sell-off, so this page records the explanation as reported and relies on none of it.

6. What it shows about restaking risk

Our guide on receipt-token discounts explains why an LRT can trade below the value of what backs it. This is the mechanism by which that discount stops being a paper loss. A holder who intended to wait out a discount, and who had borrowed against the receipt, did not get to decide whether to wait. An oracle decided, from a pool that was emptying partly because of the liquidations the oracle was causing.

The exit liquidity guide makes the related point that a receipt's secondary market is not a redemption. This event shows the sharper version: before redemption exists, the secondary market is not merely the only way out, it is also the only thing that can be used to value the asset — so the thinness of that market sets the terms for everyone who borrowed against it, including the holders who never touched it.

Nothing here involved slashing, an operator fault, or a contract defect. Both venues report their systems working as designed, and both still cost their users money. Of the three case files published here, this is the one where no component failed at all.

Sources

Every figure above comes from one of the following. Dates are publication dates, and these links open in a new tab.

1. Gauntlet LRT Core Vault Market Update 4/24/2024 (ezETH price volatility) — Gauntlet, in the Morpho governance forum, 24 April 2024. Source of the timeline, the block height, the liquidation and insolvency figures, the pool list and the liquidity drain. Written by the curator that selected these markets and whose depositors absorbed the loss.
2. Gearbox: Modular Leverage — Gearbox, 29 April 2024. Source of the Gearbox ledger, the pause side-effect, the reimbursement vote and the argument for redemption-based oracles. A party to the event; the piece is also an argument for Gearbox's own design.
3. ezETH FAQs — Renzo Protocol documentation. Cited only for the present state of ezETH redemption, not for anything about April 2024.
4. Renzo's ezETH depegs to $688 following end of airdrop farming window — Cointelegraph, April 2024. Cited as an example of the single-venue price figures discussed in section 5, which this page does not rely on.
5. Renzo increases airdrop allocation slightly after community outrage — The Block, 25 April 2024. Cited for the reported allocation change in section 5, which this page also does not rely on.

Searches for an account of the event published by Renzo returned none. That is a gap in the record rather than a finding, and it is the reason this file is scoped to the lending markets.

This page is explanatory and not a recommendation. Naming the organisations above records a documented event and reproduces their own published statements; it is not an allegation of wrongdoing and not a safety assessment.