Same disclosure as on the mWIN Horizon thread: no commercial relationship with Securitize, Neuberger Berman, Aave Labs or LlamaRisk. I mark illiquid private positions for a living and I read these as a stablecoin supplier would.
Worth saying first that this proposal answers most of what I asked over there before anyone asked it. The liquidator gate is pre-launch rather than post-launch, the mismatch between a 24-hour lock-up and a multi-day high-yield unwind is disclosed rather than buried, and the Executive Action problem is raised by the issuer rather than by a commenter. That last one is the most important section in the post and I have not seen it picked up.
The backstop number is the one I would push on. The proposal sizes liquidator backstop capital at 3-5% of borrowed TVL against a three-business-day liquidation window. That is a clear, falsifiable commitment, which is more than most proposals offer, so it deserves to be tested rather than accepted.
The risk that backstop absorbs is not a fraction of borrowed TVL – it is the NAV move across the liquidation window plus the T+1 settlement leg, against whatever headroom the LLTV leaves. The disclosed risk figures are monthly: worst month -18.25%, 2022 maximum drawdown -13.20%, and -14% to -18% for a 400bps spread widening. A liquidation window is four business days, not a month, so none of those speaks to the exposure actually being covered.
The figure that speaks to it is a four-day one, and there are two of them. Over the same July 2016 to July 2026 blend: what is the worst four-business-day fall? And what is the worst four-business-day fall that begins only after the blend has already dropped far enough to put a position into liquidation? The second is the one that sizes the backstop, because liquidations do not begin on quiet days. They begin part-way into a bad run, and bad days cluster. Sizing a backstop at 3-5% of borrowed TVL is defensible if those numbers are small. If the worst four-day move inside March 2020 was materially larger than five per cent, the backstop is sized below the event it exists for. The blend is already built, so both figures should be cheap to produce.
There is also an open thread on precisely this question. GBQuant’s independent liquidation-capacity work, posted a day after this proposal, models liquidator warehousing economics explicitly – funding cost, capital hurdle, hedge cost, execution loss and redemption throughput – and arrives at a required bonus rather than assuming one. It does not cover RWA collateral or permissioned liquidators, and it asks whether the risk framework should supplement instant clearance with a specified exit horizon and stressed primary-redemption throughput. This proposal is a direct instance of that question: permissioned liquidators, no secondary market, exit only by NAV redemption through the fund. The two threads belong together, and a 3-5% figure derived on that basis would be far more convincing than one asserted.
The two live RWA onboardings have oracles asymmetric in opposite directions. Here, a CAPO-style adapter caps NAV growth at 15% APR while downside passes through immediately with no smoothing or floor. On the mWIN thread, a static band sits 0.8% below the issue value and 15.4% above. One design bounds the direction that cannot hurt a lender and lets the other run; the other bounds the direction that can hurt a lender very tightly and lets the upside run.
Both are defensible in isolation. I do not think both can be the house view. For an asset whose own disclosure puts a -18% month inside the sample, the absence of any downside floor is a deliberate choice with a real consequence: a single inaccurate NAV print marks the entire market down at once, and every position that breaches its threshold enters the same three-day liquidation window at the same moment. Is the intended rule that RWA feeds bound manipulation upward only, and that downside protection lives entirely in the LLTV and the liquidation bonus? If so, stating it once in the framework rather than per asset would settle a whole class of repeatable questions.
On Executive Actions, the open side of the ledger. The proposal asks for AIP-level acknowledgement of how bad debt from an Executive Action is allocated between the Fund, the liquidator and the protocol, and proposes suspending accrual for frozen positions so a borrower under a legal freeze does not accrue debt they are prohibited from servicing. Both are right, and both sit on the borrower’s side of the ledger. The stablecoin supplier is on the other: funding capital that is illiquid, un-liquidatable and no longer earning, for a duration set by a court rather than a market. Does the proposed allocation contemplate compensating that supplier, and out of what?
Three smaller things, none of them risk questions. The oracle provider is named twice and differently: the primary feed is described as a Chainlink HINC NAV feed wrapped with LlamaGuard dynamic bounds, while the external dependency list names Redstone as the NAV feed, and the Chainlink feed requirement is separately noted as unresolved. Which is it, and does the answer change the attestation chain?
The bounds here are described as dynamic, where the mWIN feed’s are static values set at deployment. What makes them dynamic, what is the input, and who holds the authority to move them?
And the fund’s term is given as commencing 18 August 2026 in one place and 20 July 2026 in another – a month either way matters more than usual here, since operating history is the one thing this collateral has none of.
None of this is opposition to onboarding, and the disclosure standard here is above the norm. The backstop figure is the single item I would want derived properly rather than stated before this reaches a vote.