Aave 200M rsETH Bad Debt Event: Where the Impact Is and How It Gets Handled
Within 72 hours of the KelpDAO exploit on April 18, Aave V3 mainnet accumulated $177-200M in bad debt — the largest single-event bad-debt episode since 2020. Almost all of it sat in one market, rsETH-WETH; USDC, ETH, wstETH and other pools were essentially untouched. Why rsETH, and why did Aave’s risk machinery not function as intended?
How the Bad Debt Formed: Three Steps
Step one: bridged rsETH and mainnet rsETH decoupled. The 116,500 rsETH minted out of thin air landed on Arbitrum, but Aave’s accepted collateral was mainnet rsETH. The two should stay in price lockstep via arbitrage. But once bridged supply was being dumped, arbitrageurs grew unsure whether the same vulnerability extended to mainnet and refused to absorb flow. From 04:30 UTC, mainnet rsETH also depegged, bottoming at 0.83 ETH.
Step two: Aave’s price feed lagged. A composite oracle combining rsETH/ETH and Chainlink ETH/USD assumes deviation stays within 5% for under 30 minutes. After 05:00 UTC the assumption broke for four hours. The degradation safeguard at 05:34 locked rsETH at 0.95 ETH rather than the market’s 0.71-0.83.
Step three: liquidation routing broke mechanically. Liquidators rely on the Curve rsETH-WETH pool and DEX aggregators. Oracle at 0.95, market exit at 0.78 — every $1M processed cost $180K. Walls of below-threshold positions sat with no rational liquidator. Bad debt accumulated quietly for ~16 hours before Aave governance dropped rsETH LTV to zero and paused borrows.
Where the Debt Actually Sits: Concentration Was Worse Than Expected
Risk DAO published an on-chain analysis on day five. Largest positions:
| Borrower (last digits) | rsETH collateral | WETH borrowed (ETH eq.) | Bad debt gap |
|---|---|---|---|
| 0x…a3f2 | 28,400 | 27,200 ETH | 6,720 ETH |
| 0x…c81d | 19,800 | 19,000 ETH | 4,680 ETH |
| 0x…7e0a | 14,500 | 13,900 ETH | 3,420 ETH |
| 0x…b6c4 | 11,200 | 10,800 ETH | 2,640 ETH |
| 200+ smaller accounts | ~38,000 total | ~36,500 ETH | ~9,200 ETH |
Top four addresses produced 60%+ of bad debt. Almost identical footprints — early March, all four sold rsETH PT on Pendle, redeposited cash in Aave, used rsETH as collateral to borrow WETH and loop. Textbook LRT looping strategy — 30%+ annualized at 4-6x leverage, extremely sensitive to price stability.
Background: Aave lending basics, Aave V4 hub-and-spoke — V4 was partly designed to isolate this concentration risk.
Does the Safety Module Get Slashed: The Question on AAVE Holders’ Minds
Safety Module is the last line of defense — up to 30% of staked AAVE/ABPT can be slashed for unrecoverable losses. AAVE fell 14% in 24 hours, pricing in a slash.
Governance landed gentler. AIP-512 proposed a three-tier waterfall:
- Tier one: Aave protocol reserves absorb $68M
- Tier two: KelpDAO’s rsETH-R compensation offsets ~$72M
- Tier three: remaining $32-60M gap amortized against 18 months of revenue — no SM slash
Passed 89.3% on May 9. SM not slashed — short-term good news for holders; cost is 30-40% of distributable revenue forgone over 18 months.
Real Impact on Borrowers and Depositors
rsETH borrowers:
- Collateral oracle-locked at 0.95 ETH until April 28, no self-liquidation
- After April 28, top four flagged “unrecoverable” — positions closed, WETH debt written off
- Smaller addresses (<5,000 rsETH) force-liquidated May 6, average 18% loss
WETH depositors:
- April 18-May 6, withdrawals rate-limited to 100 ETH/day per address
- Deposit APY spiked 2.3%→11.7% from utilization distortion
- No principal lost, liquidity risk was material
rsETH depositors without active borrows:
- Partially covered by rsETH-R, active claim required
- 23% of eligible addresses unclaimed as of late May, deadline July 18
Aave’s risk machinery did its job on asset isolation; but liquidation incentives fail in LRT depeg. V4 needs to address that.
Forward Look: Is There a Future for LRT Collateral on Aave
Late May AIP-521 formalized an “LRT asset tier framework”:
- All LRT (rsETH, ezETH, weETH) reclassified Tier 2 — LTV 70%→50%, liquidation threshold 75%→60%
- “Depeg circuit breaker”: LRT deviating from ETH 3%+ within 30 minutes triggers borrow pause, repayments still allowed
- At least two independent price feeds, with conservative fallback when any diverges 2%
- Curve/Balancer LP positions no longer accepted as collateral — only underlying LRT
Capital efficiency for LRT looping drops substantially — 6x leverage maxes at 3-3.5x, yields 30%+ → 15-18%. Healthier on a risk-adjusted basis.
Full context: restaking real risks, KelpDAO exploit postmortem, DeFi guide.
Three Long-Term Questions the Incident Left Behind
First, the actual capacity ceiling for LRT collateral. Aave’s rsETH cap was 500,000 ETH; just 320,000 broke the system — no haircut for tail scenarios.
Second, liquidation incentives need redesigning under depeg. Static 5-8% bonus covers normal slippage but burns capital at 17% depeg. Needs dynamic bonus scaled to depeg magnitude — not yet formalized.
Third, the boundary between third-party insurance and SM. rsETH-R played insurance here, ad hoc. Next exploited protocol may have no treasury, SM carries full weight.
For any LRT-leveraged user, anything above 3x is almost certain to liquidate in a bridged-LRT exploit. That is the concrete number this incident leaves.