USDe Risk Audit / Lending and integrations
The lender’s price is not the holder’s exit
Two deployed cases show how valuation, debt and concentrated supply transfer risk between borrowers, vault investors and stablecoin lenders.
The pricing, position and integration observations below remain dated 17–18 September 2026. The added 20 September recovery evidence addresses cash access; it does not refresh those integrations’ installed settings.
A USDC supplier can bear USDe-dependent loss without owning USDe. Lending adds debt, collateral valuation, liquidation and withdrawal rules. A vault adds allocation and authority. The selected cases are Ethereum Aave with direct sUSDe collateral and a Base Morpho USDe-collateral / USDC-loan market reached through the Steakhouse High Yield USDC Edition vault. They are identified cases, not an exposure-weighted audit of every integration. Case selection and evidence.
The installed inputs—not the token label—set liquidation value
Aave: share accounting and USDT are the observed price inputs
At Ethereum block 26,005,206 on 18 September 2026, Aave’s installed source for sUSDe was 0x42bc86f2f08419280a99d8fbea4672e7c30a86ec. It used the sUSDe conversion ratio multiplied by a capped USDT/USD input. The sampled ratio was 1.248747741329377471 USDe per sUSDe, and the separately read quote input was 0.99923534 USD per USDT. The supplied integer calculation reproduced the accepted price of 1.24779287 USD per sUSDe. Component observations were not all one atomic batch. Exact Aave input chain.
A USDe spot discount does not necessarily change this price while share accounting and USDT remain unchanged. That can prevent forced liquidation of recoverable collateral during a temporary local discount. But a real off-chain impairment or a blocked sale can reduce liquidation proceeds without immediately appearing in those inputs. The supplier—not just the borrower—therefore matters to the assessment.
The separately dated eMode category observation used 90% LTV and a 92% liquidation threshold. Under fixed debt and no other changes, the supplied example gives 2.173913% accepted-collateral-value headroom. It is not automatically a USDe spot-depeg trigger. The base reserve configuration is different from eMode, and the investigation did not audit every debt feed or borrower account. Configuration dates and leverage assumptions.
The ratio-growth cap and the downstream USDT upper cap constrain particular upward valuations. Neither promises a yield, a par-value floor or recognition of every backing loss. A reported feed’s age alone was not converted into a heartbeat violation; the complete freshness policy was not established.
Base: par selection, a different backup unit and a real source mismatch
The exact Base Morpho market is 0x54cf9be57fdfa6457a660991907434ff9d295c465a603a50126ff647d50b7354. Core returned USDC as loan asset, remote USDe as collateral, oracle 0xf4b17c79492d68775e22e8dd0a2bb22854a39a47 and 91.5% liquidation loan-to-value. This tuple was bound through the deployed core, not only an application label. Market identity and scale.
The observed primary was configured at one USDC per USDe and selected at the sampled state. The backup used a USDe/USD feed, with no USDC/USD denominator installed in the observed configuration. Thus an independent move in USDC creates a denomination issue. Morpho’s raw 10²⁴ price scale here corresponds to one human-unit USDC per USDe, given 18-decimal collateral and 6-decimal loan assets. Constituent inputs and output units.
The 16-hour setting is not a verified safe-switch promise. The deployed oracle returned a deviation different from the candidate source’s same-input average-denominator formula. Current selection, inputs, duration getters and output are observed; full challenge, healing, fallback and initialisation behavior are not verified. A matching alternative formula at this input does not prove the whole implementation. The mismatch is source/version uncertainty, not an established exploit. Reproducible counterexample.
Par-oriented valuation can spare borrowers liquidation during a temporary discount, but it can also leave suppliers exposed to impaired collateral or local bridge-outage recovery. A later price switch would have to be assessed under the actual implementation and funded liquidation route, not an assumed timer guarantee.
A concentrated lending claim is not ready withdrawal cash
Concentration has two different denominators
The named vault is 0xbeeff2490feffa212fac2f6553682c219e6a8845; its supplying adapter is 0x4a70dc70cef79f51cc7f79c08f863b2841311165. Core, adapter and parent-vault bindings were checked. At Base block 51,477,892, market borrow utilisation was 90.080852%, with a nominal unborrowed difference of 36,389,722.650953 USDC. That difference is not vault maxWithdraw, cash reserved exclusively for the vault or a stressed recovery guarantee. Accounting, caps and concentration evidence.
The borrower must repay USDC to free collateral. Suppliers depend on idle funds, repayments and sufficient liquidation proceeds. At 91.5% LLTV the supplied Morpho formula gives a liquidation incentive of approximately 1.026167—a 2.6167% bonus, not 8.5%. High utilisation and concentrated supply can make withdrawal timing important even before final bad debt is known.
Do not infer issuer circular funding from branding. A separate Ethena x Steakhouse USDC vault was traced through its actual adapter. That adapter had 39,158 nonzero dust supply shares in this market at Base block 51,478,251, with expected assets rounding to zero USDC base units. This is counterevidence to material funding through that observed vehicle—not literally no position, proof of beneficial ownership, or proof that no other issuer-funded lender exists. Named-vehicle comparison and ownership limit.
An accepted collateral value does not fund a creditor
The 20 September recovery investigation asks the next question after the valuation maps above: who can obtain the repayment asset before the deadline? A USDC lender needs usable USDC, not simply an oracle that continues to value collateral near par. A borrower using outside USDC can repay before collateral release and interrupt the feedback chain; an unfunded borrower cannot spend future staking or fund proceeds first. Holder and creditor consequences.
The selected Maple Institutional share claim was calculated at 303,218,574.812566 USDC, while direct USDC at the pool was 694,920.022526, at Ethereum block 26,019,059, 20 September 2026, 13:35:23 UTC. The difference is not loss or total withdrawable liquidity. It identifies dependence on other assets and release conditions. The assessed wallet’s legal ownership/agency link remains unproved. Position-specific recovery evidence.
This Maple position is not the Base Morpho market or proof that the same issuer dollars finance the USDe-collateral borrowing shown above. Likewise, the Base JAAA balance is a fund-token holding—not Base USDC ready for a liquidation. Neither a network name nor a provider label connects the ownership and payment chains. The separately dated Ethena-oriented Morpho adapter dust result still has only its narrow 18 September meaning. Attribution and title limits · JAAA token versus settlement asset.
Temporary bridge funding can address timing only when eligible, funded and received in the required asset before the deadline. Associated debt or assigned proceeds remain in the ledger. The JTRSY/Basin announcement does not establish a current Ethena/JAAA backstop. No funded Base liquidation route is certified by this investigation. Conservation and bridge conditions.
Remote tokens, venues and wrappers add their own failure boundary
The Base token owner/controller and selected peer relationship were observed. They do not certify installed messaging verification, source-lock / supply reconciliation, emergency controls or a funded local conversion facility. Local bridge or conversion delay can hurt a Base USDC creditor even if Ethereum backing retains value. Invalid remote supply is a different economic-loss branch. Remote supply and access boundary.
For Bybit’s described Unified Trading Account framework, the published 100% initial-margin-ratio transfer boundary differs from the 100% maintenance-margin-ratio liquidation boundary. A transfer can be blocked first. The actual USDe-specific UTA tiers and account capacity remained unread; an Institutional Loans announcement belongs to a different product and is not substituted. Venue documents, dates and limits.
PT and YT add wrapper-specific maturity, output and income claims. No complete active pre-maturity PT market was established in the selected deeper cases. A senior or principal label is not enough to infer dollar repayment or extend these findings to every wrapper. Wrapper coverage boundary.
The stabilising route is real but conditional: funded debt repayment can replenish lenders and reduce exposure. It must arrive before withdrawals and must not rely on the same collateral that remains locked awaiting repayment. Neither inevitable contagion nor unlimited arbitrage follows from the connections alone.