DIA Oracle Mispush — IMF USDS Vault

Vault Under Review

Vault

IMF USDS · IMF-USDS

Address

0xdef1Fce2df6270Fdf7E1214343BeBbaB8583D43d

Loan asset

USDS · 0xdC035D45d973E3EC169d2276DDab16f1e407384F

Total assets (now)

819,551.68 USDS

Markets in allocation

26

Oracle provider

DIA · 0xE1A3d58dc6C516Ef18628ce7E13cfE44B4Ac1552

DIA gas wallet

0xBEE10964D1E12eE60cce1680be2a6eD955e5FaE3

Incident Timeline

• T0 2026-05-05 16:46:59 UTC — DIA gas wallet 0xBEE1…FaE3 submits tx 0x738f860b6ed20d60… to DIAOracleV2, calling setMultipleValues with 16 keys.
• T0 Block 25030092 — All 16 feeds update atomically. 16 of them resolve to a value ≥10× lower than the prior reading; every catastrophic feed is off by exactly the same factor (~10,000×) — diagnostic of an upstream decimal shift, not a market move.
• Block 25030092 — first downstream Morpho Blue Liquidate event lands. Liquidators are repaying single-digit dollars of USDS to seize entire collateral positions at the catastrophic on-chain price.
• +2h 17m — DIA's next push for MOG/USD restores it to $1.5608e-7.
• +2h 17m — DIA's push for the last affected feed (MOG/USD) restores it to $1.5608e-7. All markets back online; further bad-debt accrual stops.
• Block 25030485 — last Liquidate event in the analysis window. 184 total events emitted on Morpho Blue against vault markets.
• Now (post-mortem authored) — across the affected markets, $654,573.81 of USDS was written off as bad debt, socialised pro-rata across all suppliers. The IMF USDS vault's pro-rata share of that — its actual loss — is $33,282.75 (5.08% of the market-wide figure).

Per-Feed Price Impact

One row per key updated by the push. Prev is the last reading before T0; Bad is the value the push wrote; Restored is the first value within 5× of Prev after the push. USD figures match the protocol's own oracle.price() reading at liquidation time (decimals calibrated per-feed at d=12).

Affected Morpho Blue Markets

Markets in the IMF-USDS allocation graph, mapped to the feed they read. Highlighted rows had liquidations in the incident window.

Vault Exposure to the Bad Debt

Per-market breakdown of the vault's pre-incident supply share and its pro-rata slice of each market's bad debt. Shares are sampled at block 25030091 — the last block before the bad push.

The vault's loss is concentrated in markets where it was the dominant supplier (BITCOIN, IMF, APU, CULT). In markets where it held no pre-incident position, its share is zero and the bad debt sits with other suppliers.

What Happened

At 2026-05-05 16:46:59 UTC DIA's gas wallet 0xBEE1…FaE3 submitted a single setMultipleValues call to DIAOracleV2 (0x738f860b6ed20d60…) writing 16 keys at once. 16 of those keys came in roughly four decimal places below the previous on-chain reading — a uniform ~10,000× compression, identical across every affected feed regardless of asset class (wstETH, cbBTC, BITCOIN, PEPE, MOG, SPX, IMF, JOE, CULT, REKT, NPC, NEIRO, SHIB, APU). That uniformity rules out market movement; it points to a unit/scaling mismatch upstream of the DIA contract.

Every Morpho Blue market backed by an affected feed reads its price through an oracle wrapper that passes the underlying DIAOracleV2 value through. The moment the push landed, those wrappers returned prices ~10,000× lower than a block earlier, and every previously-healthy borrower in the affected markets crossed their LLTV in the same block.

Across the 184 liquidations that followed, liquidators paid $157.9299 total in USDS to seize $1.72M of borrower collateral (priced at the post-rectification reading). The $654,573.81 shortfall — debt that couldn't be covered from collateral at the broken price — accrued as bad debt across the affected markets.

DIA's next push for the affected feeds, landing 2h 17m later, restored prices to their pre-incident range. By that point the liquidations had already executed.

Methodology

Source data: Ethereum mainnet via eth_getLogs + eth_call against a QuickNode endpoint. DIA OracleUpdate events for the affected feeds were pulled from a ±7,200-block window around block 25030092 (capped at chain head — the incident is recent, so the post-bad-push window is bounded by current head).

Per-key decimal counts are first inferred against published reference prices, then calibrated by reading each market's Morpho oracle wrapper price() at head and back-deriving the decimals that make the post-rectification DIA value match the protocol's own canonical reading. Result: every feed lands at d=12. Cross-validation: the decimals back-derived from oracle.price() match the live DIAOracleV2.getValue() reading to 0.00% drift on every feed except wstETH (0.51% — staking-ratio jitter).

Liquidations are pulled from Morpho Blue's Liquidate events between the bad-push block and chain head, filtered to the vault's 26-market allocation graph (sourced from blue-api.morpho.org). Sample liquidation logs were re-decoded directly from chain and verified to match stored fields byte-for-byte. Collateral USD impact uses each market oracle's own price reading at head; loan-token USD uses unit-USD for USDS (verified $0.9998 via Morpho API).

Vault exposure: for each affected market, Morpho.position(id, vault).supplyShares and Morpho.market(id).totalSupplyShares are read at BAD_BLOCK − 1 (the last clean state). The vault's loss in that market is supplyShares / totalSupplyShares × Σ badDebtAssets across that market's liquidations. Summed across markets, this gives the vault's true pro-rata exposure of $33,282.75 — versus the $654,573.81 market-wide bad debt that's socialised across all suppliers in those Morpho Blue markets.

Validator: src/validate.ts · 35 / 35 checks pass. Generated 2026-05-06T05:01:49.476Z. Replayable from src/fetch.ts in this directory.

Liquidation Log

All 184 Morpho Blue Liquidate events that fired against the vault's markets between block 25030092 and 25030485, ordered by block. Repaid is the USDS the liquidator paid to take the position; Seized is the fair-market USD value of the collateral they walked away with (priced at the protocol's own post-rectification oracle reading); Bad debt is the market-wide write-off; Vault's share is this vault's pro-rata slice of that write-off.