Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Default and workout

The question an allocator asks last and cares about most: the borrower misses a payment — then what, exactly, and when do I see cash?

This page gives the sequence. Where a step is contracted it says so; where it is intended but not yet arranged it says that instead, because the difference is the whole value of the page.

The timeline

WhenWhat happensWho actsStatus
Day 0Scheduled payment not received. The settlement is simply never recorded — no receipt is written, so NAV does not moveThe registry, by omissionImplemented. settlement_never_premature and the receipt design mean a missed payment cannot be booked as anything else
Day 1–5Grace and cure period; obligor notifiedIssuer / security agentFacility documents — terms not published here
Day 5–15Event of default declared; acceleration considered. Monitoring firm confirms the collateral pool's stateSecurity agent, monitorContracted; the monitor is appointed by the insurers and paid by the borrower
Day 15–30Claim notified to the insurer. The 90-day clock starts here, not at day 0IssuerPolicy — and see the caveat below
Day 15–45Enforcement over the collateral pool: metal realised, noteholder ranking seniorSecurity agentContracted
Day 30–60Realisation proceeds distributed through the waterfall — holders first, capped at parWaterfallImplemented and proved (frt_distribute_holders_first_par_capped)
Day ~105–135Insurance settles, if the policy responds and the claim is admittedInsurerSee the gap below
BeyondResidual recovery, disputes, cross-border enforcementCourts / arbitrationUnmodelled and jurisdiction-dependent

The liquidity hole, stated precisely

Between the missed payment and the insurance settlement there is a window of roughly three to four months in which the loss is real, the claim is pending, and no insurance cash has arrived.

Insurance improves recovery. It does nothing for liquidity in that window. During it, a redeeming holder is paid from the treasury buffer, the secondary pool, or realised metal — not from the policy.

This is the single most under-appreciated feature of the structure, and the model is explicit that it does not solve it: the 90-day delay is carried as a parameter and then proved to be read by nothing (eval_arch_ignores_unmodeled_params). It is excluded, not mitigated.

Where the insurance is thinner than the timeline implies

The day-105 row assumes a policy that responds to the holder. On the documents as they stand, three things qualify that:

  • the first-tranche policy names a different insured and a USD 10m insured amount, not the programme;
  • policy rights cannot be assigned without the leading insurer's approval, and loss-payee status needs an approved endorsement;
  • the deck's cut-through — direct recourse to the reinsurer — is therefore not established for a holder today.

Until an endorsement exists naming the intended beneficiary, model the insurance layer as a recovery to the named insured, not to you. See the first facility.

What realisation actually looks like for gold

Gold is the favourable case, and it is worth saying why: it is fungible, priced continuously, and sells into a deep market without needing a buyer for a specific asset. That is a genuine structural advantage over most private-credit collateral, where liquidation can be theoretical for want of a bid.

Three frictions remain, and none is modelled:

  • Assay and form. Doré is not bullion. Realising it may require refining first, which takes time and costs money.
  • Location and control. Realisation speed depends on where the metal sits and whether release instructions are actually controlled — the distinction risk and collateral treats as central.
  • Haircut versus liquidation value. The advance is set below collateral value, but the model has no calibrated haircut field; the borrowing base's deductions are uncalibrated. Recovery experience would set them, and there is none.

What is proved about the default path

More than one might expect, but narrowly.

PropertyTheorem
Protection cannot pay more than was loststacked_protection_le_loss
The same loss cannot be recovered twiceno_double_recovery
A write-off requires the recovery to be resolved firstwriteoff_requires_recovery_resolution
Default does not imply the claim is completedefault_does_not_imply_completion
Holders rank first at par in liquidationfrt_distribute_holders_first_par_capped
The distribution conserves the pool exactlydistribute_conserves

What is not proved is timing. Every theorem above is about amounts and ordering. The clock — grace, notification, waiting period, realisation window — is outside the model entirely, and it is where the loss actually gets felt.

What would close the gap

A liquidity facility or a committed backstop sized to the waiting period; an endorsement establishing the holder's insurance rights; a published realisation policy with an expected window for gold; and a calibrated haircut set against observed liquidation experience rather than assumption. None of these exists today.