Collateral value is not repayment cash

Start with the obligation, not the collateral label

A borrower owes an asset to a lending system. The protocol values the pledged collateral to decide whether the position is adequately secured. A supplier, meanwhile, owns a claim on the loan asset—not a direct right to spend the borrower’s collateral. A vault can put another allocation and withdrawal layer between the supplier and that lending market. Those relationships determine whose liquidity or equity is affected first.

Accepted collateral value is the amount recognized under the integration’s chosen inputs. Releasable collateral is the portion that can leave while the remaining position and account satisfy their conditions. Timely repayment cash is the correct debt token actually available at the paying account before the obligation falls due. A favorable number in the first category does not prove either of the others. S01 · S07 · S08

The selected cases span Ethereum Aave V3 direct sUSDe collateral, the existing Base Morpho USDe/USDC market and its concentrated vault, and an active Ethereum PT-sUSDE-26NOV2026 lending market. These are consequential identified cases, not a ranking of the world’s largest exposures. The Base borrower was selected from recent recorded activity; the Aave account used later is expressly hypothetical.

The new financial reads refresh only the explicitly dated objects below. They do not close the source-matching, exceptional-oracle or remote-supply questions in the September 26 control continuation, and they do not turn the issuer’s Maple or JAAA recovery claims into money available to an individual borrower.

Base: gross collateral substantially exceeds the no-repayment release bound

At Base block 51,851,358, 27 September 2026, 06:54:23 UTC, the sampled position held 3,038,825.552941956775445158 USDe against a calculated 2,704,559.828740 USDC obligation. Its accepted price was 1 USDC per USDe, with 91.5% loan-to-value liquidation threshold. The resulting financial health factor was 1.028087954: above one at that snapshot, not an observed liquidation.

The address was identified by a bounded Borrow-event sample, not selected as the largest or most distressed account. The included block governs the calculation; the service’s outer block number was one later. Its two sampled idle token balances were zero, but that reveals neither the owner’s identity nor other wallets, assets, credit lines or willingness to contribute capital. Selection, identity and recorded inputs

Debt accounting comes before a release estimate

Morpho borrow shares are not USDC units. The calculation uses the published share convention, including one virtual asset, one million virtual shares and upward rounding to raw USDC units:

Debt in raw units = ceil[borrow shares × (total borrow assets + 1) / (total borrow shares + 1,000,000)].

The sampled account’s borrow shares were 2,669,447,754,655,058,898. Its market update time equalled the included chain time in this batch. That avoids mixing an older market accrual timestamp into this particular calculation; it does not make the resulting debt an executable repayment quote at a later block. S09

The financial bound is necessary, not an instruction to withdraw

For one collateral and one debt, let Q be collateral units, D be debt units, O be accepted collateral value in debt units, L the liquidation threshold, and h the chosen remaining health factor. Holding all inputs fixed:

No-repayment release bound: max(0, Q − hD / (L × O)).
Repayment required before releasing x: max(0, D − L × O × (Q − x) / h).

At h = 1, the bound is 83,022.461423 USDe. At an illustrative h = 1.01 it is 53,464.430508 USDe. A one-million-USDe release requires at least 839,034.447798 USDC of repayment under the first calculation. Releasing all collateral in a repay-first sequence requires the full calculated debt.

Neither bound is a measured withdrawal limit. Contract rounding, permission checks, accrued interest before inclusion, reserve conditions and execution can tighten the result. Exact equality at health factor one is not asserted to be executable or prudent. The calculations correct two opposite mistakes: debt does not necessarily lock every unit, but a healthy position does not make the entire collateral balance free to sell. S07

At fixed debt, a decline of approximately 2.7320575% in accepted collateral value reaches health factor one; at fixed collateral value, debt growth of approximately 2.8087954% does so. By contrast, selling all collateral would cover current principal before costs at about 0.890001674 USDC per USDe. That last number is a gross economic recovery threshold—not proof that all collateral is releasable, that this sale is funded, or that an available liquidator can realize it.

The earlier Base market batch returned a par-valued oracle and primary selection. It did not verify future challenge, healing, stale-input handling or a funded local backstop. A normal returned price and a healthy accounting result do not establish the exceptional protection on which a creditor might rely.

Aave: shares in the reserve are not a borrower’s free USDC

The Ethereum 26,067,090 snapshot returned about 140.4 million sUSDe shares at the a-sUSDe receipt address, with almost the same receipt supply. That is reserve-wide staking collateral. It is not a particular borrower’s executable withdrawal, nor the USDC or USDT a loan-token supplier needs.

The category bitmaps include collateral reserve IDs 30 and 32 and borrowable IDs 3 and 8; the reserve records identify USDC as 3 and USDT as 8. The installed sUSDe source address matches the earlier source. Later reads identify sUSDe as its ratio provider and the existing base quote adapter, with isCapped=false at that sample. Multiplying the first batch’s staking ratio by its USDT price reproduces the accepted sUSDe price after integer scaling. This is a financial consistency check, not a new source-code, cap-policy or freshness audit. Recorded observations

Headroom depends on the price the account actually uses

For an assumed existing eMode-2 account, 79,992.067266 sUSDe shares represent 100,000 USDe at the sampled ratio. Against 90,000 USDC debt, accepted collateral value is 99,989.642482 USDC and health factor is 1.022116. The slight difference from exactly 90% LTV comes from the actual relative accepted prices. This is not a measured Aave wallet or proof that a new loan of that amount could be originated.

Under fixed collateral quantity and threshold, the financial sensitivity is:

New health factor = old health factor × accepted collateral-price factor / (debt-price factor × debt-quantity factor).

A 5% recognized collateral-price decline gives 0.971011 in this example. A 3% rise in USDC’s accepted dollar value gives 0.992346 with the collateral input unchanged. A 3% decline in the USDT input transmitted to collateral gives 0.991453 with USDC pricing unchanged. None is a prediction or a threshold for an arbitrary USDe spot discount. A USDT borrower may experience some offset when both sides use that input, but caps, feeds, timing and the whole account still matter.

Current variable-borrow rates in that batch were 4.3035919115% annualized for USDC and 5.4759461234% for USDT. Holding them fixed solely for simple one-day arithmetic, 90,000 units accrue about 10.611596 USDC or 13.502333 USDT. These are not future rates, APYs or offered refinancing terms. Debt growth can consume headroom while a holder waits.

Loan-reserve cash is a separate resource

These balances are not reserved for USDe borrowers and are not issuer reserves. The receipt and debt-token quantities use the corresponding underlying denomination; their separate decimal getters were not refreshed. This is not an exhaustive accounting reconciliation: direct cash plus sampled variable debt exceeds aToken claims by 230,687.163715 USDC and 248,568.515304 USDT, respectively. Treasury accrual and other adjustments were not reconciled. The differences are not labelled missing assets or solvency surpluses.

An independently funded USDC repayment adds USDC before the borrower’s later staking sale. A same-asset receipt-token repayment or a collateral conversion can have a different resource effect. Aave’s published withdrawal and repayment descriptions recognize both account collateral constraints and available reserve liquidity; neither can stand in for the other. S01 · S02 · S03

Base suppliers: a concentrated accounting claim is not reserved cash

At Base 51,851,215, 06:49:37 UTC, the selected vault’s sole installed adapter again identified the vault, Base USDC and the same Morpho core. The market’s supply accounting was 389,552,488.710380 USDC against 350,805,225.729610 USDC borrowing. Their nominal difference was 38,747,262.980770 USDC, with stored utilization of 90.053391%.

The market’s stored time was 24 seconds earlier than this batch. The expected-assets getter can use a different accrual convention; it is not forced into equality with a fraction of the stored market total. Both concentration ratios retain their own denominators. Compared with the September 18 selected-market fraction of 88.761411%, the new fraction is lower while the expected-assets amount is larger. That is a dated comparison, not a reconstruction of flows or proof that risk fell. Recorded observations

Nominal unborrowed assets are not an exclusive allocation to the vault, a measured maxWithdraw, or an aggregate withdrawal quote. Other suppliers and borrowers can use the same resource. Other vault positions may be liquid, but their accounting value cannot be credited as timely cash without examining those routes. S08

A timely repayment helps once, not twice

To compare the actual borrower with a withdrawal deadline, the example uses the later borrower batch, not the earlier vault batch. At Base 51,851,358, the market’s nominal unborrowed assets were 38,772,553.686476 USDC. Assume a 50-million-USDC withdrawal from this market, no other withdrawals and no other funding credited. This is not observed demand or a total-vault withdrawal estimate.

The improvement is exactly the borrower’s 2,704,559.828740-USDC calculated debt, only if independently funded and received in time. Repayment converts a receivable into cash. It does not create a second asset. Cash paid to an exiting supplier is unavailable for another payment. Nor do these figures establish Ethena ownership: the earlier separate issuer-oriented adapter had only dust in this market, and no new beneficial-ownership graph is supplied.

When loans remain recoverable but repayment dates follow withdrawal requests, suppliers may face delay without principal impairment. When collateral recovery leaves residual debt, the outcome changes to loss. An adapter holding almost all market supply shares concentrates that exposure for its vault; that economic conclusion does not certify the complete deployed loss-accounting implementation.

One Ethereum sale route behaves very differently at different sizes

The identified Ethereum USDe/USDC pool is 0x02950460e2b9529d0e00284a5fa2d7bdf3fa4d72. Its own getters returned the pair’s canonical tokens, name and coin-accounting balances. At Ethereum block 26,067,144, 06:56:47 UTC, its independent price views returned:

The pool’s own accounting balances were 250,885.086884337938879366 USDe and 243,146.029994 USDC. They were not separately queried ERC-20 balances. The get_dy views do not execute a swap, consume liquidity, establish a minimum amount received or measure transaction inclusion costs. The outputs cannot be summed as four sequential fills. Pair, method and raw-unit record

The 100,000-unit result provides a close-to-nominal-par indication at that size. It does not scale tenfold. The one-million-unit result is a limit of this particular route and snapshot, not a market-wide USDe depeg. Neither output establishes future liquidity after a cooldown, the best aggregate route, issuer redemption, or Base USDC proceeds.

This is an additional debt-asset-aligned Ethereum route, not a revision of the older thin USDe/USDT route. The September 18 pool-specific quotes retain their pair, size and dates. Different routes can produce different results without either observation being erased.

The old creditor can be paid while a new creditor remains exposed

The following sequence is funded by explicit assumption, not observed funding. It is not the sampled Base borrower or a measured Aave account. It combines the Aave snapshot’s share ratio and category inputs with the separately timed 100,000-USDe quote, held fixed as a hypothetical later sale. Those observations are not an executable transaction package.

Assume an existing eMode-2 Aave position with roughly 79,992.067266 sUSDe shares representing 100,000 USDe, and 90,000 USDC debt at repayment. Assume the account is permitted to repay, withdraw and enter the staking exit, and that its later claim can be released. Crucially, a distinct capital provider supplies 90,000 Ethereum USDC before the old debt is repaid. This is not money borrowed from the same locked collateral without an established route, an issuer reserve transfer or an unverified PSM advance.

The assumed term credit costs 10% simple annual interest, actual days/365. These are illustrative terms, not offered credit. Gas, additional service fees and conversion costs are excluded. They would reduce the borrower’s residual.

The measured 86,400-second setting is only one part of that sequence. It does not bound detection, repayment, collateral release, transaction inclusion, claim execution or later conversion. The inherited staking mechanics remain those of the earlier claim analysis; the new read refreshes the duration, not every restriction or withdrawal condition.

One day is an idealized timing illustration with immediate surrounding processing, not a promised or established service time. If the term loan is due at day one but sale proceeds arrive at day three, the earlier obligation still lacks funding unless another arrival or an agreed extension covers it. A positive eventual residual cannot finance an earlier deficit on its own.

For the one-day case, the incremental USDC ledger is +90,000 to Aave, −99,923.537191 to the pool/buyer side, +24.657534 net interest to the term lender and +9,898.879657 to the borrower, apart from display rounding. The initial and returned 90,000 principal offset for the lender. This is a transfer ledger, not a complete balance sheet or creation of backing.

The larger same-route case does not fund the new lender

In a separate assumed example, one million released USDe follows 900,000 USDC of initial repayment funding. At the same one-day credit convention, 900,246.575342 USDC is due. Applying the one-million-unit pool output solely as a same-route sensitivity yields 243,084.492699 USDC. 657,162.082643 USDC of additional cash is required for full repayment.

Without that contribution the new creditor is not paid in full. A balanced ledger that shows all obligations paid must explicitly include the missing contribution; it must not disguise the shortfall as an acceptable negative borrower cash balance. This is route-size risk, not a forecast of issuer insolvency or a proposed sale.

Other financing routes can change the sequence—but need evidence

Full-debt prefunding belongs to this repay-first, release-all, wait-and-sell sequence. Partial free collateral, a funded buyer of free sUSDe, an available collateral-conversion facility or a permissible refinance can change the timing. No sized sUSDe sale or executable atomic refinance was established here.

A conventional flash loan must settle inside its transaction and cannot remain unpaid through an ordinary cooldown. A permitted transaction that instead opens new debt replaces an obligation; it does not eliminate credit. A router name or product description alone does not identify sufficient funding, acceptable output and a payment deadline. S01 · S04

Liquidation eligibility is not a profitable, funded liquidation

A third party repaying debt needs collateral proceeds sufficient to cover that payment and its costs. Let O be the accepted collateral value used to size liquidation, q the realized proceeds per collateral unit at the relevant size, and k the incentive factor. For potential repayment R, the simplified gross sale margin is:

Gross margin = R × (k × q / O − 1).

Using Morpho’s published formula and the observed 91.5% threshold gives k = 1.026167265… and a gross break-even condition q/O ≥ 0.9745. Gas, financing and actual execution costs tighten that condition. It is not a 2.55% liquidation trigger, quoted exit, safety score or parameter recommendation. Eligibility, financing and realization remain different questions. S07 · S15

Each case assumes one million USDe, 930,000 USDC debt and 91.5% LLTV. These normalized positions already lie beyond the financial threshold; they are not asserted to be newly originatable at that threshold or to describe the actual account. No inferred account history, contract transaction, integer edge-case test or realized loss is supplied.

The first case shows why total collateral worth more than debt can still fail to attract a funded liquidator. The second recognizes a lower value and transfers additional collateral away from the borrower; a later price recovery does not automatically reverse that transfer. The third represents reduced final recovery, with the incentive itself consuming value otherwise available to creditors.

A par-oriented mark can protect a borrower during a recoverable local discount while a supplier waits for cash. A market-responsive mark can recognize impairment sooner while forcing a borrower into a thin sale. The correct comparison asks which party benefits, at whose expense and with what funded exit. None of these alternatives is selected as the verified future Base response.

Denomination matters too: a one-dollar asset buys about 0.970874 USDC when USDC is worth $1.03. The inherited Base backup lacked a USDC/USD denominator, while Aave separately values dollar debt inputs. That arithmetic illustrates different exposures, not a forecast or a fresh exceptional-feed assurance. A bridge interruption or unverified local route also prevents importing an Ethereum sale into a Base cash ledger.

Active November PT: its accounting unit is not a cash payout

A principal token (PT) separates a maturity-linked principal claim from the yield-token claim. Ordinary reads identify PT-sUSDE-26NOV2026, its standardized-yield wrapper (SY) and yield token (YT), positive supply and an Ethereum Morpho USDC lending market. This is an active identified maturity, not a recycled expired token or a presumed Aave PT listing. Its expiry is 26 November 2026, 00:00 UTC, approximately 59.7104 days after the September 27 identity snapshot. S11 Recorded observations

The SY rate is measured in USDe accounting units per SY and matched the same-call sUSDe conversion. Under a purely constant-rate normal-maturity illustration, one USDe accounting unit corresponds to about 0.799919696 sUSDe. The future ratio can differ. “One USDe worth of sUSDe” must not become “one sUSDe,” and neither phrase supplies USDC. S12

The route for an unencumbered PT holder passes through the appropriate maturity/claim processing and SY redemption into sUSDe, then a free-share sale or staking withdrawal into USDe, and finally a funded conversion if USDC is required. For a borrower, the collateral-release condition comes first. Maturity changes the collateral claim; it does not pay or forgive the separate USDC loan.

The market record read on September 27 had a stored last-update time of 25 September 2026, 20:03:47 UTC. Its 738,859.634500-USDC nominal unborrowed difference is therefore not presented as a freshly accrued cash balance. The 7,868,589.982965234368322354 PT held at the entire Ethereum Morpho core also spans all markets and cannot be assigned wholly to this one.

Principal is conditional even in accounting units

Pendle’s documentation describes a below-watermark exchange-rate case that reduces principal recovery and interrupts yield accrual until recovery. The observed SY rate was above the sampled stored PY index, not evidence of a present negative-yield loss. A separate illustrative watermark of 1.25 and later rate of 1.20 gives 0.96 accounting units per normalized claim. That is a financial example, not a predicted or tested outcome for this series, and not an extra haircut to add to the same underlying loss twice. S13

A pre-maturity sale could avoid waiting until November but would have its own price, size, financing and settlement conditions. No exact Pendle AMM binding and sized pre-maturity exit were established. The new result closes the narrower absence of an active identity/output comparison, not the whole PT exit question.

What the evidence now supports—and what remains partial

The useful advance is financial and position-specific: a real Base debt/collateral sample; refreshed selected reserve and market inputs; borrower/supplier distinctions; a time-ordered unwind with explicitly assumed outside credit; and a concrete active principal-token output comparison. This is not an observed end-to-end unwind or a complete Aave-versus-Base funded-recovery assessment.

Patient unlevered holders, leveraged borrowers, loan-token suppliers and outside term lenders bear different outcomes. A borrower can lose equity before long-term collateral recovery. A supplier can wait without a principal loss or face residual bad debt after genuine impairment. A new creditor can inherit the remaining realization risk after the original protocol is paid.

Timely outside repayment, accessible free collateral, other liquid vault assets and a funded local sale are genuine stabilizers when their conditions hold. No issuer reserve, Maple/JAAA claim, PSM seed or documentary bridge promise is added as an independent cash source here. A repayment reduces a receivable; a cross-chain transfer changes location; a bridge loan adds debt. The same resource is counted once.

The control investigation remains partial, but that is not the only reason the integration work is partial. The missing financial routes and measured account evidence are independently important. Current reserve adequacy, probabilities of loss, investor suitability and trade recommendations are outside the established result.

Evidence, dates and numerical inspection

Financial/configuration observations are provider-mediated read-only responses. Eight separately timed batches include block/time getters; two natural Base Borrow logs were used to select the example. These are financial observations, not independent security tests. No transaction receipt, executed sale or completed funded unwind was established.

Public-source descriptions were retrieved for the research on September 27. Undated descriptions are not given an invented publication date. On-chain timestamps are not measured HTTP retrieval times, and separate batches are not one cross-chain balance sheet. Calculations and assumptions are labelled separately below.

For the Base borrower the outer service reported 51,851,359; the included 51,851,358 identifies the data used. The preceding Base market’s stored time was 24 seconds before its batch. The PT lending record stored a September 25 update despite being read on September 27. None of these distinctions is silently corrected by using the chapter date.

Download selected numerical observations and scenario inputs (CSV). This is a reader selection with units, dates, provenance and limits, not the full source record or an additive balance sheet. Preserve addresses and long decimals as text on spreadsheet import. The earlier September 20 reader data remains unchanged.

The locators above identify the dated records, not future settings or beneficial ownership. For the 18-decimal collateral and 6-decimal loan tokens here, Morpho’s raw oracle value of 10²⁴ means one loan token per collateral unit after its 10³⁶ scaling.

Public source trail

The following sources support the dated findings. Retrieval on 27 September 2026 does not replace a document’s publication date, a record’s economic cutoff or an observation’s block.