TL;DR

How it works

PYUSD deposits mint senPYUSDmain shares; NAV grows as borrower interest accrues. Vault V2 (unlike MetaMorpho V1) routes through a single adapter (0xc2A1…34d4, a Morpho Market V1 adapter → the immutable Morpho Blue singleton 0xBBBB…FFCb), risk-bounded by an abstract cap system, not a market queue. 8 markets carry a funded position (verified by direct position() reads; the eval previously said 9). Review reports 10 markets hold non-zero caps, the two unfunded being a PT-sUSDE market and WETH — the cap count is not independently re-derived here, so treat “8 funded” as the verified figure. Live allocation (2026-07-21): kBTC 29% · syrupUSDC 21% · weETH 15% · LBTC 15% · sUSDe 11% · cbBTC/WBTC/wstETH ~1.2% · idle 8%. The “PayPal USD” label reads like a plain stablecoin product; the exposure is a leveraged DeFi-credit book — ~45% BTC-wrappers, ~32% yield-bearing stables, ~15% ETH-LST. Drift since the July review is rotation within the same book, not a change of character: kBTC and LBTC up, sUSDe down, no new collateral type added.

Oracle setup. The vault only lends, so it is never itself liquidated; its NAV absorbs bad debt when a borrower’s collateral gaps below the loan before liquidators clear it. Every market quotes collateral in PYUSD/USD (Chainlink): kBTC off BTC/USD with no kBTC-specific feed (a kBTC depeg is invisible); syrupUSDC/sUSDe off the collateral’s ERC-4626 NAV × peg feed; weETH/LBTC off exchange-rate feeds (LBTC’s base leg Redstone).

Yield. One leg enters NAV — borrower interest, spot supply ~2.5–4% by market (kBTC 2.73%, syrupUSDC 3.96%, weETH 2.50%, LBTC 3.02%, sUSDe 2.83%, 2026-07-21), blending to a 2.77% gross / ~1.75% net of the 1% management fee. Demand is utility-driven — loopers financing a carry, not emission-rented TVL — but reflexive: rates track borrow utilization (AdaptiveCurve IRM), not a risk premium, so the flagged-riskiest market (kBTC) still pays below syrupUSDC. Read the headline carefully: the ~4.8% quoted netApy is ~3 points of external reward distribution on top of a ~1.75% in-NAV return. Rewards are a separate distributor stream — they never touch share price, can stop without notice, and are not what a redemption pays out. Size on the 1.75%.

Risks

Verified on-chain (2026-07-21): allocation, LLTVs, utilization, oracle feeds, current fees, owner/curator roles + thresholds, per-selector timelocks (incl. setIsSentinel/setCurator = 0), gate abdication, forceDeallocatePenalty, the sentinel set (isSentinel — the curator is one of the two), and the 2026-07-13 kBTC cap-raise event decoded from its raw log. Carried forward unverified from 2026-07-07 (14d): the allocator key set. Taken on trust: Paxos reserve attestation, Sentora’s allocation intent, audit scope, the Vault V2 fee/penalty ceilings (5% / 50% / 2% — source constants whose getters revert), kBTC/syrupUSDC/sUSDe issuer solvency.

#RiskMechanismSource + our delta
1kBTC oracle blind spot + 29% concentrationLargest market (0xe51f…eef5, 29% of TVL, up from 27%) prices kBTC purely off Chainlink BTC/USD — re-verified 2026-07-21: BASE_FEED_1 = “BTC / USD”, QUOTE_FEED_1 = “PYUSD / USD”, no kBTC leg. A Kraken-custody kBTC depeg is invisible to liquidations at 86% LLTV → bad debt, unpatchable (markets immutable). Exit: kBTC↔BTC spread >200bps >1h.On-chain oracle read; NOCA-original. kBTC is a newer wrapper than cbBTC — thinner track record.
2Recursive yield-collateral credit at 91.5% LLTV — and it loops back to a position we may hold directlysyrupUSDC (21%) + sUSDe (11%) = ~32% of TVL: PYUSD lent against tokens tracking Maple’s mark-to-model NAV / Ethena’s USDe peg. A write-down or depeg gaps past the oracle’s ERC-4626 read → bad debt. The syrupUSDC leg is sharper: this vault is 61% of all syrupUSDC-collateralised Morpho lending, priced off Pool.convertToAssets() — Maple’s own NAV, which only ever rises and books losses on Maple’s schedule. No early warning is available from this leg, and a markdown hits here and any direct syrupUSDC holding at once. Exit: syrupUSDC/USDe deviation >2% for >24h; treat a Maple markdown as an immediate bad-debt event here.Maple credit risk per our maple-syrupusdc eval; Ethena short-vol carry. Surface traced on-chain in the syrupUSDC collateral surface study — NOCA-original.
3Single-key curator/owner + instant fee seizureOwner (0xe8C9…8008) and curator (0x9e39…aac0) are each 1-of-1 Safes of a bare EOA. setManagementFee/setPerformanceFee have 0 timelock → 5% + 50% instantly; setIsAllocator instant. Yield seizure, not principal. Exit: any fee raise.On-chain thresholds + timelock reads. Curator = Sentora — mitigant is prompt exit, not governance.
4⚠ FIRED — curator raised the kBTC cap; and the “2 sentinels” barrier is largely illusoryNew-market caps and new adapters carry a 3-day timelock (259200), after which the curator can allocate principal. This trigger has already fired inside the review window: at block 25,525,822, 2026-07-13 19:33 UTC (tx 0x54125e59…acce6), an Accept(bytes4,bytes) of increaseAbsoluteCap raised the kBTC market’s absolute cap to $120M — the largest and (per Risk #1) riskiest market. Verified from the raw log: topic1 = 0xf6f98fd5, value 0x6d23ad5f8000 = 120e12, collateral 0x73E0C0d4… (symbol() = “kBTC”), LLTV 86%. Neither the prior eval nor this refresh caught it until review. The stated barrier does not hold: isSentinel returns true for exactly two addresses, and one of them is the curator Safe itself, so a compromised curator already owns half the veto; the only independent sentinel (0x74D1…1618) is an EOA with nonce 0 — it has never sent a transaction. And setIsSentinel and setCurator both carry 0 timelock, so a compromised owner deletes both sentinels and installs itself as curator instantly, leaving no revoker at all. Adapter registry is still frozen (Morpho-vetted types only). Exit: any further Accept/Submit or adapter event — and note the cap headroom now exceeds the current $94M kBTC position.On-chain event log, isSentinel, and per-selector timelock reads (2026-07-21).
5Bad-debt gap-down tail at high utilizationMarkets run ~84–91% utilization with no loss reserve; a fast BTC/ETH crash can outrun permissionless liquidations at 86–91.5% LLTV. Compounds with thin instant liquidity (Risk #6).Standard Morpho Blue lending tail; on-chain utilization 2026-07-21.
6Liquidity-gated exit~$59.6M instant (18.4% of TVL): ~$27M idle + ~$33M unborrowed — improved from $51.3M / 15.9% in July, driven by the larger idle balance. The other ~82% needs borrower repayment or forceDeallocate (permissionless in-kind exit, ≤2% ceiling). That penalty is no longer zero: forceDeallocatePenalty = 1e14 = 0.01% — economically trivial, but it confirms the lever is live and raisable on the 3-day timelock. Exit: instant ratio <10%.On-chain 2026-07-21. Tighter than a queue-based MetaMorpho — the vault is ~100% of supply in each market.

Key-compromise threshold

Morpho Blue’s core is immutable; withdrawals cannot be frozen (3 of 4 gates permanently abdicated, the 4th gates only deposits) and the adapter registry is frozen. Privilege concentrates in two single-signer Sentora Safes.

Liquidity & exit

Primary path is ERC-4626 redeem at NAV. totalAssets() = 323,536,423 PYUSD, idle PYUSD.balanceOf(vault) = 26,998,337 (2026-07-21, block 25,582,684). Instant capacity = idle + Σ min(vault position, market unborrowed) across the 9 markets ≈ $59.6M (18.4% of TVL); the remaining ~$264M depends on borrower repayment, fresh supply, or forceDeallocate (an in-kind exit that itself needs market liquidity, so no free lunch under stress). No secondary AMM market for senPYUSDmain.

Defensible size at current liquidity (post-haircut): $15M. That is ~25% of the $59.6M instant depth and clears same-day at NAV; under a 3× stress haircut (→ ~$19.9M available) $15M still fits inside a 1-day window. We size to instant liquidity, not the $324M TVL, and hold at $15M despite the improved depth — the constraint that binds is Risk #2, not depth. Stress scenario (“NOCA exits 50% in 7d while peers redeem”): a $7.5M exit against ~$20M post-haircut is comfortable; a full $15M coincident with peer redemptions and a utilization spike >98% would queue the residual for borrower-repayment cycles. Aggregate with any direct Maple position (Risk #2) rather than sizing the two separately. Halve if not refreshed by 2026-10-21 (90d).

External reviews

No independent risk rating (LlamaRisk / Chaos Labs / Bluechip / DeFiScan / Exponential) of this specific Vault V2 was located — it is new (deployed 2025-12-16); Staking Rewards lists only a yield profile, and Sentora, the curator, is a conflicted source (discount per rubric.md). Refresh before relying; per rubric.md, >90d (⚠️) is stale.