The markets
On 25 August 2026 two Morpho lending markets on Ethereum lent against the same collateral: PT-reUSD-10DEC2026, the principal token on reUSD, a product of Re (re.xyz). One lent USDC, the other USDT. Both had a liquidation threshold (LLTV) of 91.5%, and both valued the PT through the same oracle.
The trades
Between 04:28:47 and 04:37:35 UTC, one address made eleven trades on the Pendle market for that PT, each buying YT with SY. Together they spent 320,000 SY-reUSD, about $351,000 at the SY's own on-chain exchange rate that day. Buying YT through the pool is selling PT into it: the trades sold 5.43 million PT into the pool (more YT came from limit orders filled on the way), and the PT's implied rate rose from 10.58% before the first trade to 20.37% after the last.
What the oracle read
The markets' oracle takes the lower of two prices for the PT: a 15-minute time-weighted average of the Pendle market's rate (Pendle's TWAP oracle, over 900 seconds), and a fixed curve at a 6% yearly discount. All morning the fixed curve stayed above the TWAP, so the TWAP priced every liquidation.
A TWAP follows the market with a lag. At 04:37:47, twelve seconds after the last trade, the oracle price stood 0.54% below where it had been before the trades, and the first loan was liquidated. By the last liquidation in the USDC market, at 04:48:35, it was 2.45% below; by the last in the USDT market, at 04:51:23, 2.76%.
The liquidations
- USDC market: 16 liquidations of 13 borrowers, 35,188,279.71 USDC repaid, from 04:37:47 to 04:48:35 UTC.
- USDT market: 17 liquidations of 7 borrowers, 955,966.41 USDT repaid, from 04:39:23 to 04:51:23 UTC.
- One borrower was liquidated in both markets, so 19 borrowers in all. Neither market recorded any bad debt.
Liquidators who sold the seized PT into the same pool carried its implied rate further, to 22.43% at 04:45:35. The three largest liquidations, which seized 32.95 million PT between them, did not sell into the pool in the same transaction.
Why so small a move was enough
A loan with a 91.5% threshold can be liquidated once its health factor (collateral value × 0.915 ÷ debt) falls below 1. For the oracle's path that morning to reach them, every borrower liquidated in the USDC market must have started below a health factor of about 1.025, which is leverage of roughly 9.3 times or more. At that leverage, a fall of about 2.4% in the oracle price is the whole margin.
Morpho's contract lets a liquidator take collateral worth the debt repaid times an incentive factor set by the threshold. At 91.5% that factor is 1.0262, a bonus of 2.62% of the debt repaid, taken from the borrower's collateral. The PT seized in each liquidation matches that figure.
What moved was the PT's market rate. The oracle followed it, as a TWAP is built to, and the most leveraged loans had less room than the move.
Method
Every figure is read from the chain: the Liquidate events of the Morpho Blue contract for the two markets, and the swap and rate events of the Pendle market, through Blockscout's public explorer interface, each decoded and checked again. The markets' ids were derived again from their creation parameters. The TWAP was rebuilt from the Pendle market's rate events, and it reproduces the price behind every liquidation to within a millionth.
References: the USDC market on Morpho Blue, id 0x1e9d614631a7df0ec07fb05b2c8cb2491575fd1a63a33bf187a6afb295a4fc64; the USDT market, id 0x6acd18854e43f93f5a90446004c35c72a68af1ad99f08cb247782ce6457914ca; the Pendle market, 0x13285bcbc27f92b47b4edb99d744c07b48c977c0. The note cites these and no transaction or account.
Limits
- No borrower's own health factor before the event was read. The 9.3 times is a bound derived from the oracle's path, not a measurement of each loan.
- USDC and USDT are counted at one dollar each.
- The size of the two markets before the event was not read, so the share of each that was liquidated is not stated.
- Borrowers are counted, not named, and the address that traded is not named either. Every transaction behind these figures is public on the chain; the note names none of the people in them. The record shows what happened, not why.
- Both legs of the oracle run Pendle's published oracle code; neither was deployed from the factories in Pendle's published list, and who deployed them was not established.