IPOR Fusion Liquity ETH carry

Our funds can get stuck when:

• The operator cannot complete the unwind: insufficient BOLD or failed trades leave collateral locked.
• Liquity’s collateral rules block release: even having enough BOLD did not guarantee closure in the tested price stress.
• Withdrawal processing stalls: competing requests consume released capacity, requests expire, or the operator does not act. We resolved the first two in simulations.

Position and current exit evidence

Withdrawal pin: Ethereum block 25,990,633, 16 September 2026, 14:27:59 UTC. Targeted withdrawal sequences were added September 17 at that same archived state. NOCA is the largest shareholder with 42.08% of the vault and a 2,213.52 WETH gross claim. Only 262.98 WETH is idle across the vault. With the operator processing our request first, we received 2,208.68 WETH and cleared our shares. After all other holders were paid first, the operator’s remaining unwind paid us 2,200.21 WETH, leaving 3.97 WETH unpaid.

A combined run was more restrictive: after all 125 other holder requests were paid and outside Curve LPs exited, the remaining LP quoted only 32,040 BOLD against 2.836m BOLD debt. The BOLD route failed. An existing USDC exit route and collateral withdrawal raised our recovery from 101.00 to 1,988.75 WETH, leaving a 250.83 WETH claim behind an open loan at 121% collateralization. That fallback is a conditional test, not a recommended operating level or a maximum recovery bound.

The same treasury Safe owes 2,166.16 WETH to Morpho against 1,900 wstETH, but its latest borrowing funded both a 1,000 WETH vault deposit and a 1,166 WETH purchase of liquid wstETH. It is not solely a Liquity funding loan.

Without an unwind, we recovered 262.57 WETH if first and zero after another holder consumed the available cash. The original full-vault-unwind benchmark also repaid Morpho and recovered all collateral; the new targeted tests leave that financing unchanged. A sustained Morpho lender run plus operator delay made the combined loan liquidatable in the 30-day simulation. The detailed withdrawal investigation covers ordering, requests, underlying Curve liquidity, controls, liquidation, branch shutdown and repayment from separate treasury assets.

Historical economics below: 19 August 2026 inputs, not a current yield quote. This refresh updates withdrawal and funding evidence; subsidy funding and net carry have not been refreshed.

The provisional annualized net-yield estimate is 6.51% after management and performance fees but before the 0.40% entry/exit spread and unwind costs. A discretionary weekly WETH transfer contributes 5.89 points, with 1.593 WETH left in the payer wallet against a 1.74 WETH four-week average payment. Without it, the estimate falls to 1.21% versus 2.185% for holding wstETH. Depositor funds are controlled by a 2-of-3 Safe of bare EOAs with no timelock, and no incorporated operator or named principals were verified from the materials reviewed.

Liquity’s team separately communicated to NOCA that it will maintain 6% net APY after fees for total vault NAV up to 5,000 ETH through Q1 2027 and cover NOCA’s 20bp entry fee. That commitment is not encoded in the vault or visibly escrowed. The 20bp exit fee and unwind costs remain.

Mechanism and true exposure

WETH enters an ERC-4626 vault, is converted to wstETH and posted to a Liquity V2 trove. The trove mints BOLD, which is deployed in a Curve BOLD/USDC LP and staked in its gauge, with a small Uniswap V4 position also present. At the September 16 pin, the vault held 4,013.90 wstETH collateral, owed 6.740m BOLD at 0.75% annual interest, held 57.05% of Curve pool LP supply, and kept 5.00% of its NAV idle. The name still does not describe the actual collateral: this trove holds wstETH. The treasury’s Morpho borrowing adds a financing layer outside the vault; vault shares and Morpho debt cannot be cancelled internally. Exiting requires actual WETH to reach Morpho.

The August oracle review identified two vault oracle paths. IPOR NAV uses Chainlink-based WETH/wstETH inputs and a Pyth BOLD adapter; the adapter calls getPriceUnsafe, and the deployed manager does not enforce its returned publish time. Liquity liquidation separately uses Chainlink STETH/USD multiplied by the wstETH/STETH exchange rate, with a one-hour heartbeat, 24-hour staleness threshold and ETH/USD/exchange-rate fallback (Liquity source, accessed 20 Aug 2026).

Capital, return, control and exit paths

Solid petrol lines are assets or exits; dashed petrol is debt; teal is return; coral is the vault-specific subsidy. Dotted lines are fees, control or dependencies.

Historical vault-only diagram: block 25,789,465, 19 August. Its 15% buffer and subsidy funding are historical; current idle cash is 5%. The external Morpho financing leg is described above.

Net economics, benchmark and capacity

At the historical August 19 rates, staking yield minus BOLD interest plus Curve trading fees is −0.14% of NAV; CRV and gauge-wide BOLD incentives raise the pre-subsidy return to 1.84%. The 5.89-point WETH subsidy and 0.50% management plus 10% performance fees produce the provisional 6.51% estimate. IPOR Labs receives 0.30% management and 2% performance; Sentinel receives 0.20% and 8%. The estimate annualizes point-in-time staking, borrowing and reward inputs; it excludes the 0.40% entry/exit spread and unwind costs. The operator app showed 6.03%, and the difference remains unreconciled. A fixed 90.4 WETH annual subsidy dilutes as NAV grows; the supporting table shows that sensitivity. The Liquity-team commitment changes the commercial return floor through Q1 2027, but it does not explain the funding mechanism or make the observed weekly subsidy durable on-chain.

5.9 of the 6.5 yield points are a discretionary subsidy

Source: Ethereum block 25,789,465, 19 Aug 2026. Lido staking rate from its live API on the same date; CRV emissions use the verified 1.0× rate. The operator app showed 6.03%; the 0.48-point difference remains unreconciled.

Exit, cost and latency

The operator must obtain BOLD, repay Liquity, release collateral and sell it for WETH. There is no verified service-time guarantee.

The combined shareholder/Curve run blocked the remaining BOLD exit even with a cooperative operator. Using the existing Curve exit fuse to receive 2.957m USDC, then selling it and 534.54 wstETH of permitted collateral, paid 1,988.75 WETH in total. 250.83 WETH remained. The loan still owed 2.836m BOLD; collateral withdrawals were constrained by its minimum ratio. Other BOLD sources and the final payout remain unverified. The same outside-Curve shock without prior shareholder exits allowed our smaller exit to complete, so sequencing changes the result.

Requests charge 0.20% in shares and expire 24 hours after creation. A delayed operator had 2,358.75 WETH ready, but the expired request and a direct redemption both failed while cash was reserved. Renewing the request and release paid 2,204.40 WETH, including a second request fee. Separately, a USDC freeze of the Curve pool blocked the top-up route, but buying BOLD through extra LP redemption allowed our full request to be paid. A failed route does not establish that every route fails.

With no Curve shock, processing all other requests before ours paid us 2,200.21 WETH, leaving a 3.97 WETH claim. Full-share redemption requires a 2% liquidity allowance plus reservations; small assets and reserved cash can prevent the last shares exiting.

Current Liquity ICR is 1.7701, giving 32.21% price headroom to the 120% liquidation threshold at fixed debt/collateral. This is distinct from Morpho’s 1.0531 health factor: its exchange-rate oracle has approximately 5.04% downward headroom, and its debt grows while the vault is illiquid. The full-collateral stETH sale quoted 0.0546% below input at the pin; 0.1638% is the mechanical three-times-slippage floor. Comparable historical stressed slippage and guaranteed 1/7/30-day exit capacity remain UNVERIFIED. Exact quantities, assumptions and fork receipts are in the withdrawal investigation.

Risks

IPOR later announced a separate OpenCover wrapper; this memo evaluates the uncovered base vault (IPOR announcement, accessed 20 Aug 2026).

Unresolved conditions

• Supply an attributable explanation and authorization record for the 18 August EIP-7702 delegation.
• Identify the contracting entity and principals, with a signed curator agreement and recourse terms.
• Demonstrate that fuse and oracle changes require delay or independent approval and that operational keys are not bare EOAs.
• Verify the deployed vault/fuse version against its audit scope; require a BOLD publish-time guard and alert on Liquity oracle fallback or branch shutdown.
• Document the Liquity-team 6% / 5,000 ETH / Q1 2027 commitment, the 20bp entry-fee reimbursement and its funding mechanics; place the required subsidy under visible funding.
• Reconcile the operator’s 6.03% return with this 6.51% estimate and document performance-fee crystallization and any high-water mark.
• Demonstrate executable 1-, 7- and 30-day exit capacity for the contemplated ticket under concurrent redemptions, operator delay and the worse of 3× current or comparable historical stress slippage.

Trust and control

The current administrator Safe has three bare EOA owners and a two-signature threshold. Two keys sufficed for the tested custody drain and pause. One Alpha key can act within the configured fuses and determine unwind/release timing; one stale fee key can alter request economics. This is not a proof that trading permissions cannot cause loss with fewer than two keys. The IPOR DAO Safe is 4-of-6; both treasury Safes are 3-of-6 with the same EOA owners. The observed module lists were empty. All four nonce-filtered Safe queues returned zero pending transactions when checked in this session.

Evidence boundary

The September 16 withdrawal investigation supersedes the older exit amounts and adds current ownership, financing, controls and local transaction execution. It preserves its pin, source hashes, eight original withdrawal tests and thirteen protocol stress scenarios. The September 17 targeted-exit evidence adds two staged withdrawal sequences and one smaller-exit Curve stress at the same pin. Six operator-processing sequences additionally test competing releases, late completion, a combined withdrawal run with fallback recovery and its no-shock control, and a USDC freeze. Current APY, subsidy durability, historical stressed slippage and operator response commitments are not refreshed. The monitor records direct-exit, control and funding-attribution gaps explicitly; it does not establish Tower readiness. Its share-balance exposure lifecycle alone cannot establish repayment of Morpho.

The following paragraph describes the historical August economics evidence only.

Vault balances, roles, signer composition, delegation state, Liquity positions, redemption order and reward transfers were measured at Ethereum block 25,789,465; subsidy transfers cover mid-May through that block. The annualized yield is a point-in-time model, not a realized holding-period return. The Lido rate and public operator, IPOR and coverage statements are external inputs dated and linked above. The 6% net-return and entry-fee terms are a Liquity-team communication to NOCA, not an on-chain obligation. The deployed-code audit delta, BOLD freshness control, fee crystallization, operator-rate reconciliation, delegation purpose, legal recourse, forward subsidy funding and stressed exit remain unresolved. The D2, Vega-Lite and Great Tables sources beside this memo reproduce the exhibits.

Supporting evidence

Subsidy durability

The payer delivered 14.167 WETH from mid-May through the pinned block. The most recent seven weekly payments totalled 10.53 WETH; the latest four averaged 1.74 WETH. The payer wallet held 1.593 WETH. Over nine funded weeks, 51.5% of the depositor subsidy came from Liquity incentives and 48.5% from Sentinel cash.

Weekly WETH payments exceed the funded balance

Source: Ethereum block 25,789,465, 19 Aug 2026. Payer: Sentinel’s Atomist Safe and a contract it controls.

Capacity if the subsidy budget stays fixed

The operator stated that it intends to hold net yield near 6%, which would require increasing the WETH budget as the vault grows. The table isolates the observable floor case: 90.4 WETH per year spread across a larger NAV, net of the 0.40% round-trip spread. It holds CRV and gauge-wide BOLD rewards, borrow rates, market impact and the 15% idle policy constant, so it measures subsidy dilution, not executable strategy capacity.

Source: Ethereum block 25,789,465, 19 Aug 2026. Incentive held fixed at 90.4 WETH per year.