# Adversarial source review: peg defence and endogenous collateral

Review date: October 29, 2026. Inputs: the supplied report and its claims dataset. Scope: ten claims carrying the conclusions about contraction, incentive funding, endogenous backing, fee hooks and synthetic-dollar failure. This is a source-content check, not an independent transaction replay, reserve audit or review of every claim.

**Result: nine hold; one is unsourced for purposes of independent verification because its cited statements are inaccessible. No selected claim is demonstrated wrong.** “Holds” means the source supports the claim at its stated level of attribution, including appropriately labelled inference. It does not turn an issuer observation into independently verified chain data. “Unsourced” below means the exact assertion could not be substantiated from readable cited material; it does not mean the inaccessible statement never existed.

Each block reproduces the claim text exactly, excluding its inline bibliography. Source descriptions are paraphrases. **Fact** identifies what the retrieved source says; **inference** identifies my assessment; **uncertain** identifies what remains unverified. All linked pages were opened on the review date. Mutable documentation is the retrieved version, without a historical commit guarantee.

## 1. C009 — Curve v2 intervention failed when needed

**Claim as written — Fact:**

> V2’s safety gate failed operationally during the June 12, 2024 upward depeg. LlamaRisk’s contract-linked investigation found the spot/EMA deviation exceeded the permitted ±0.0005, preventing supply precisely during liquidation demand. The first successful updates in the three examined pools came roughly 525 blocks, or 45 minutes, after onset. The stronger crvUSD quote also made collateral appear cheaper in crvUSD-denominated lending markets, causing spillover liquidations. This is measured intervention latency; the report does not supply a single reliable consolidated peak price to reproduce here.

**What the source actually says — Fact.** LlamaRisk’s July 1, 2024 [incident report](https://www.llamarisk.com/research/crvusd-incident-report-20240612), under “Pegkeeper Incident Analysis,” identifies the v2 implementation introduced June 3 and three Ethereum pools: USDC, USDT and PYUSD against crvUSD. It specifies the ±0.0005 spot/EMA tolerance and reports first updates about 525 blocks after onset, approximately 45 minutes. Its summary and mitigation discussion describe additional Curve Lend liquidations and the collateral-denomination problem. It distinguishes those lending markets from aggregate-USD-priced crvUSD mint markets.

**Verdict: holds.** The date, version, chain, tolerance and latency agree. The claim correctly avoids inventing a consolidated price peak. **Uncertain:** I did not replay calls or verify every plotted block. **Inference:** this supports a warning about an intervention gate blocking legitimate demand, not a conclusion that oracle guards should universally be removed.

## 2. C015 — GHO’s actual exit depends on reserve inventory

**Claim as written — Fact:**

> The GSM gives an actual stablecoin exit: below peg, buy GHO and burn it for an approved reserve asset, less fees, while inventory lasts; above peg, deposit that asset, mint GHO and sell it. Exposure caps limit external-asset holdings; facilitator buckets constrain issuance; oracle freezes can disable conversion. Funding comes from stablecoins deposited by minters and potentially DAO capital. Early risk recommendations proposed $3.5 million per USDC/USDT module and 0.2% fees, conditioned on an already much stronger peg. These recommendations are not deployed limits.

**What the sources actually say — Fact.** Aave’s current [GSM help page](https://aave.com/help/gho-stablecoin/stability-module) describes governance-approved token conversion and the initial fixed 1:1 strategy. The [development thread](https://governance.aave.com/t/gho-stability-module-update/14442) begins August 11, 2023, but the relevant recommendation is Gauntlet’s **December 2, 2023** reply, not the opening post. It recommends 3.5 million caps for USDC and USDT and 0.2% buy/sell fees, conditional on a substantially recovered peg, including consistently above $0.995 and a preceding 30 days above $0.97. It discusses freeze bounds and reserve drainage. Earlier replies recommend different caps and zero fees; they are superseded recommendations within this discussion, not contradictory deployed settings.

**Verdict: holds, with a citation precision limitation.** The numbers are correctly described as recommendations for the Ethereum launch, not live limits. The economics of reserve-limited conversion follow. **Uncertain:** these two pages do not themselves establish a particular DAO funding transfer or live facilitator bucket. “Potentially DAO capital” is a funding possibility, not evidence of execution. The dataset’s broad date `2023-08/2024-01` should not obscure the December 2 origin of these exact figures.

## 3. C025 — StablePair is not a blanket tax on depegging trades

**Claim as written — Fact:**

> Uniswap v4 supports per-swap LP-fee overrides in dynamic-fee pools. StablePair, launched September 10, 2026, supplies a concrete Ethereum implementation for USDC/USDT and USDC/USDG. Inside its reference band it adjusts fees to quote a consistent bid and ask before price impact. Outside the band it charges zero on trades moving farther from reference and auctions the corrective trade through a decaying fee. Below and above reference the directions reverse. This deliberately captures arbitrage value for LPs; it is not a surcharge on every depegging sale.

**What the sources actually say — Fact.** The official [v4 dynamic-fee documentation](https://developers.uniswap.org/docs/protocols/v4/concepts/dynamic-fees) permits per-swap dynamic LP fees. The [September 10, 2026 launch announcement](https://blog.uniswap.org/stablepair-hook-a-fee-that-moves-with-the-market) names Ethereum mainnet USDC/USDT and USDC/USDG pools. The [StablePair fee documentation](https://developers.uniswap.org/docs/protocols/uniswap-labs-hooks/stable-pair/concepts/dynamic-fees) explicitly assigns zero fees to adverse trades outside the band and decaying fees to corrective trades. It says price is read at block start and the consistent in-band quotes exclude price impact.

**Verdict: holds.** Launch date, chain, pairs and fee directions match. The report’s pre-impact qualification is essential and correct. **Uncertain:** this is implementation/launch documentation, not a verified historical parameter snapshot or an outcome study. **Inference:** LP value capture is not proof of aggregate dollar-peg improvement; the report properly keeps those objectives separate.

## 4. C034 — sUSD failed through changed demand and exits, not an isolated ratio experiment

**Claim as written — Inference:**

> The sequence supports a demand-and-exit failure before any demonstrated exhaustion of marked collateral: change debt ownership and repayment incentives, remove routine borrower buybacks, create selling pressure, subsidize a lockup, then ultimately retire the liability. Cutting a collateral ratio increases potential issuance and reduces loss tolerance, but the evidence does not identify that numerical cut alone as the depeg’s cause. A design can remain overcollateralized on an oracle screen while its holders lack an attractive route to realize a dollar.

**What the sources actually say — Fact.** [SIP-420](https://sips.synthetix.io/sips/sip-420/), created January 6, 2025 and currently marked Implemented, describes delegated protocol-owned debt, a proposed 200% pooled issuance ratio, 1,000% solo-staker ratio and twelve-month debt transfer. Its network field includes Ethereum, Optimism, Base and Arbitrum. [SIP-421](https://sips.synthetix.io/sips/sip-421/), created April 24, 2025 for Ethereum, explicitly attributes missing sUSD demand to Jubilee participants being able to exit holdings. It adds five million SNX of rewards and locked sUSD deposits. [SIP-423](https://sips.synthetix.io/sips/sip-423/), created June 12, 2026 and marked Implemented, sunsets legacy sUSD on Ethereum and Optimism. It specifies four SNX per entitlement with a one-year lock and one-year vest, while explicitly deferring receipt-contract and staking-reform work.

**Verdict: holds as inference.** These sources support the incentive-change sequence and eventual retirement decision. They do not isolate the collateral-ratio cut as a causal variable. The claim says “before any demonstrated exhaustion,” not that a reconciled system balance sheet proves solvency throughout. **Uncertain:** realized sales, net endogenous backing and executable SNX depth at the trough remain unmeasured here. Deferred reimbursement must not be described as already completed cash redemption; this claim does not do so.

## 5. C035 — Djed preserves ADA contraction while restricting expansion

**Claim as written — Fact:**

> For Cardano’s Djed deployment, the backing asset is ADA: effectively 100% network-token reserve composition by design, not 100% issuer-governance-token backing. DJED holders have a senior claim; SHEN holders finance and absorb volatility in the surplus ADA reserve. The operator specifies a 400–800% target reserve range. Below 400%, new DJED minting and SHEN redemption stop, while DJED redemption into ADA remains available. Below peg this preserves a contraction route; above peg a mint stop can prevent supply arbitrage from eliminating a premium.

**What the sources actually say — Fact.** COTI’s [August 25, 2023 clarification](https://medium.com/cotinetwork/answers-and-clarification-to-recent-djed-community-questions-95176450db09) identifies Cardano ADA backing and a 400–800% range, with continued DJED-to-ADA burning. The [Djed Alliance mechanism page](https://docs.stability.nexus/stablecoins/djed-overview/how-the-protocol-works) describes reserve coins as claims on surplus base-coin reserves, stablecoin minting blocked below the minimum ratio, and reserve-coin exits allowed only above that threshold. Importantly, its generic redemption rule is full dollar-equivalent base coins only above 100% coverage, otherwise pro rata.

**Verdict: holds for the stated mechanism.** ADA-only reserve composition by design is not a measured portfolio snapshot. SHEN is not a second independent backing asset. The report elsewhere supplies the below-100% qualification, so this paragraph’s “remains available” should not be mistaken for unconditional par redemption. **Uncertain:** the Alliance page is generic documentation, not a dated audit of COTI deployment bytecode. **Inference:** the possible premium when minting is restricted follows mechanically, but is not itself a continuous market-price observation.

## 6. C048 — Elixir’s reassurance-to-shutdown sequence and remaining balances

**Claim as written — Fact:**

> Elixir’s November 3 statement, accessed through a mirror, asserted special one-for-one redemption rights on its Stream lending position and said it was unwinding. Its November 6 mirrored statement instead described mint/redeem shutdown and a holder snapshot. It said roughly 90% of the remaining deUSD supply, around $75 million, was held by Stream and a similar fraction of remaining backing was a Morpho loan to Stream. Those are post-run remaining-balance fractions, not the pre-crisis reserve allocation or a proof of repayment capacity.

**What the cited sources actually say — Uncertain.** Both cited mirrors — the [November 3 statement](https://teapaw.org/elixir/status/1985371198210064474) and [November 6 statement](https://teapaw.org/elixir/status/1986443495351927257) — were inaccessible through the web reader. Direct HTTP retrieval also returned 403. Attempts to open the corresponding [first original X post](https://x.com/elixir/status/1985371198210064474) and [second original X post](https://x.com/elixir/status/1986443495351927257) failed. There is therefore no readable source text from those citations for me to compare with the precise promises, dates, 90% fractions or approximately $75 million balance.

**Additional corroboration — Fact.** Elixir’s readable [official deUSD documentation](https://docs.elixir.xyz/deusd), which redirects to a deprecated-product page, confirms that deUSD is sunset, has no value, and has a separate one-for-one USDC claims process. It does not establish the November 3 promise or the November 6 concentration figures. This corroborates the end state, not every historical assertion in C048.

**Verdict: unsourced for independent verification of the exact historical statement and figures.** This is an access/provenance gap, not a finding of fabrication or a numerical contradiction. The report openly identifies mirrored access and records the missing authenticated archive in G08. An authenticated statement archive or transaction-reconciled snapshot is needed before treating the specific remaining-balance percentages as verified. The report correctly warns that remaining balances are not pre-run reserve allocation.

## 7. C054 — Resolv’s supply controls failed independently of reserve custody

**Claim as written — Fact:**

> The breached service could finalize minting without an independent on-chain maximum-output or collateral-ratio check. CertiK traces the privileged completion path; Resolv acknowledges that signing authority translated directly into unconstrained minting. Nexus Mutual reports Curve USR moving from $1 to $0.025 within 17 minutes. That is a reported pool-specific low, not a promise that every holder could trade at that price. The amount illicitly issued and the attacker’s extracted proceeds are different loss measures.

**What the sources actually say — Fact.** [CertiK’s March 23, 2026 analysis](https://www.certik.com/blog/resolv-protocol-incident-analysis) describes SERVICE_ROLE invoking `completeSwap`, a minimum-output check without a maximum, and two transactions issuing 80 million USR. [Resolv’s April 4 post-mortem](https://resolv.xyz/blog/resolv-postmortem-march-22-2026-incident), “Root Cause Analysis,” explicitly acknowledges the absence of independent on-chain constraints once signing authority was obtained. It identifies Ethereum transactions and distinguishes 80 million issued from approximately $25 million extracted. [Nexus Mutual’s March 23 account](https://nexusmutual.io/blog/resolv-protocol-incident-report-by-nexus-mutual) reports the Curve price falling from $1 to $0.025 in 17 minutes and describes missing mint/collateral checks.

**Verdict: holds.** The mechanism and quoted time/price match the sources and March 22 Ethereum incident. The price is appropriately attributed to Nexus and a particular venue. **Uncertain:** I did not replay pool events to establish a minimum independently. Nexus compresses the two-mint sequence in its opening narrative; the report appropriately uses Resolv’s more detailed timeline elsewhere rather than asserting that all 80 million were minted in the first transaction.

## 8. C059 — Ethena’s venue crash was not a uniform on-chain peg collapse

**Claim as written — Fact:**

> For October 10–11, 2025, the committee’s analysis reports USDe at 0.65 USDT on Binance and 0.92 on Bybit, while major Ethereum and Plasma DEX pools stayed around $0.99. On Binance it spent 23 minutes below 0.90, 53 below 0.95 and 86 below 0.99. Binance separately acknowledged forced liquidations from the dislocations. An exchange mark-price feedback loop was therefore a severe user-loss event even though the issuer’s on-chain collateral and primary redemption machinery did not suffer the same collapse.

**What the sources actually say — Fact.** Blockworks Advisory’s [October 15, 2025 governance analysis](https://gov.ethenafoundation.com/t/usde-analysis-during-the-oct-10th-11th-volatility-event/706) reports the Binance 0.65 and Bybit 0.92 lows, the 23/53/86-minute thresholds, and Ethereum/Plasma major-pool prices around 0.99. It discusses Binance internal-order-book pricing and reports orderly primary redemption without a backlog. [Binance’s October 11 announcement](https://www.binance.com/en/support/announcement/detail/f3d5d97ed12c4f639019419bd891705e) independently acknowledges depegs and resulting forced liquidations involving USDE, BNSOL and WBETH.

**Verdict: holds as an attributed venue-specific account.** Binance confirms the liquidation harm, not the exact minima, threshold durations or the full oracle-causation account. Those details rely on the governance analysis. Its redemption totals are internally inconsistent: the summary uses $1.2 billion against $8.6 billion supply; the dedicated section uses $1.9 billion against $14.6 billion. The report explicitly identifies that conflict in C060/G11, so it is not an unacknowledged contradiction in C059. **Uncertain:** precise flow reconciliation and independent replay remain missing. No inference of uninterrupted par on every venue is justified or made.

## 9. C066 — DeFiChain itself says fees did not repair the DUSD overhang

**Claim as written — Fact:**

> DeFiChain’s November 20, 2024 account expressly says dynamic fees and interest had been ineffective against the algorithmic DUSD overhang and inadequate collateralization. Its replacement involved a snapshot and controlled release of liquidity in 100 tranches. Even the optimistic restart account put DUSD near $0.30. It does not isolate the surcharge’s causal effect from the pre-existing balance-sheet problem.

**What the source actually says — Fact.** The official [November 20, 2024 restart account](https://blog.defichain.com/2024/11/a-successful-restart-for-defichains.html) explicitly describes ineffective dynamic interest and fees, inadequate collateralization and unstable DUSD. It describes snapshot-based limited initial liquidity and reintroduction in 100 conditional tranches, then reports DUSD near $0.30. These are DeFiChain observations, not Ethereum trades; no particular price venue or timestamp accompanies the approximate quote.

**Verdict: holds.** Date, tranche count and approximate price match. This is an issuer acknowledgement that the overall arrangement failed to restore a dollar peg, not an identified estimate of how much the surcharge itself changed price or volume. The report preserves that distinction. **Uncertain:** the source supplies neither a controlled fee counterfactual nor quantified diversion to untaxed venues.

## 10. C085 — Incentivized crvUSD liquidity proved temporary

**Claim as written — Fact:**

> Pangea’s December 30, 2025 study gives a less favorable interval: TVL-weighted mean absolute pool-price deviation rose from 0.000186 to 0.000309 across equal pre/post-YieldBasis periods. It expressly does not isolate causality from market volatility. Incentivized PYUSD pool TVL rose from about $2 million to $30 million, then returned just above $2 million as rewards tapered. Its end-of-block Ethereum analysis supplies measured evidence that incentives rented liquidity; it does not identify a surcharge effect or contradict the differently sampled August 2026 result.

**What the source actually says — Fact.** Pangea’s [December 30, 2025 study](https://blog.pangea.foundation/does-yieldbasis-hurt-the-crvusd-peg/) describes end-of-block observations for four Ethereum crvUSD pools, TVL weighting and equal pre/post-YieldBasis periods. It reports mean absolute deviation increasing from 0.000186 to 0.000309 and PYUSD-pool TVL rising from approximately $2 million to $30 million, then falling just above $2 million as incentives decreased. Its conclusion explicitly acknowledges that higher market volatility is uncontrolled. It also identifies a large October 10 end-of-block distortion as a multi-block MEV event rather than a sustained structural depeg.

**Verdict: holds as a reported observational pattern.** The numbers and caveat are accurate. “Incentives rented liquidity” is a reasonable interpretation of the reported reversal, rather than an experimentally isolated elasticity estimate. **Uncertain:** the page does not provide exact textual comparison-window boundaries or a downloadable query for independent replication. The finding cannot establish that all incentive-funded liquidity is equally transient, that YieldBasis caused the deviation increase, or that fee tolls caused migration. The report makes none of those stronger assertions.

## Conclusions that exceed, or remain unsupported by, the evidence

1. **C050’s exact three-day reassurance-to-shutdown demonstration remains unverified.** The general counterparty-concentration risk is a sound inference, and official documentation confirms deUSD’s end state. The precise November 3/6 wording and C048 concentration fractions require the inaccessible statements or an authenticated archive. They should not carry the same verification weight as the readable Synthetix and Resolv primary documents. This is already substantially acknowledged in G08; it is not evidence that the conclusion is false.

2. **No numerical debt ceiling follows from these cases.** C076’s proposed inequality `(1+b) × r × D <= eta × L(s,T)` is a modelling choice, not an empirically fitted rule supplied by the sources. Neither dependable-depth discount `eta`, run fraction `r`, stress horizon, nor an acceptable impact is calibrated. The report correctly labels this inference and explicitly declines a transferable ratio in C078/G01. Its stronger wording that a ceiling should be below stress-tested capacity is a conservative recommendation, not a result proved by the collected observations. In particular, the inequality itself constrains a fraction of debt; it need not imply total debt is below depth when `r` is small. Do not convert it into a validated issuance limit without additional assumptions.

3. **Hook recommendations are design hypotheses, not demonstrated aggregate peg performance.** C073’s preserve-corrective-arbitrage recommendation is economically plausible. StablePair documents LP-value capture and can charge corrective trades while charging adverse trades zero. DeFiChain documents an unsuccessful overall system. Neither measures how a proposed directional hook would affect a thin endogenous-collateral stablecoin across venues. The report appropriately marks the implications as inference and records the missing experiment in C069/G12. It would exceed the evidence to present these recommendations as a proven peg defence or a measured volume-retention result.

4. **Redemption availability is not sufficient proof of a durable dollar floor.** C001/C070 support making contraction possible, but successful redemption into an illiquid own token still depends on outside buyers, oracle valuation, execution costs and staking delays. Djed’s mechanism documentation is not a continuous peg series; sUSD’s experience is not an experiment proving that a particular redemption implementation would have prevented its depeg. The report generally states these limitations. A claim of demonstrated safety for the intended small-token system would not follow.

## Acceptance assessment

**Inference:** The report is useful and generally disciplined about attribution. It distinguishes governance proposals from executed limits, venue prices from global pegs, native network tokens from issuer governance tokens, and compensation from ongoing dollar convertibility. The sampled sources do not reveal a substantive numerical or version contradiction. The dataset agrees with the ten quoted report claims and explicitly retains the principal gaps.

**Uncertain / unanswered:** The brief’s historical executable SNX/ADA/TRX market depth, reconciled whole-system endogenous sUSD backing, numerical debt-to-depth calibration, causal surcharge efficacy and venue migration remain unanswered. Some current caps and deployment provenance are also unresolved. These are material completeness limits, not hidden factual answers. Acceptance as an attributable research synthesis is reasonable; acceptance as empirical calibration or independent on-chain validation is not established by this review.