This historical replay is not a forecast.

Today’s tape did not clear Seiche’s full-story gate. There was no fresh measured change, and the newsworthiness score was 5, below the threshold for a full story. This is therefore a desk brief and a historical replay, not breaking news or a forecast.

The question is whether dollar-funding capacity can tighten through the calendar and balance-sheet channel before overnight cash prices show scarcity. The April 2025 tariff-shock basis unwind supplies a conditional lesson: when a common exit meets limited dealer intermediation, Treasury selling, basis pressure and balance-sheet demand can become one transmission chain.

## The mechanism or historical setup

The April 2025 tariff-shock basis unwind followed a policy shock that changed risk and volatility faster than leveraged relative-value books could adjust. The desk guide’s lesson is not that a large position alone creates a problem. It is whether many holders need the same exit while dealer balance sheets are absorbing the same volatility shock.

That distinction matters for funding analysis. A position may be financeable in orderly conditions. But if holders try to reduce the same trade as volatility rises, dealers may be asked simultaneously to intermediate Treasury flow and absorb volatility. Financing terms, basis pressure and Treasury selling then become connected rather than separate observations.

The current setup has a different initiating force. It is a dated reserve path, not an observed shock in overnight funding. Reserves fell $49.3B in the week through 2026-08-12 to $2,944B. The Treasury General Account rebuilt $56.6B and the foreign reverse-repo pool absorbed $39.7B, while the residual released $36.5B. The reported legs sum to the reserve change, with $18.9B, or 0.28% of assets, remaining in the unnamed residual.

That identity is not proof of stress. It identifies a plumbing condition that could matter if demand for cash and dealer balance-sheet capacity arrives at the same time. The replay’s lesson is conditional: calendar drainage becomes consequential only if it meets a concentrated need to de-risk and constrained intermediation.

## Quantified evidence with as-of context

Seiche’s composite read 45.2, labelled STRAIN. It is a Seiche derivation, not an observed market price. The dated reserve path was the largest contribution: the weather component scored 100.0 and contributed 11.0 points.

As of 2026-08-13, Seiche’s Tell was +37.1. Plumbing stood at the 64th percentile of its own history while market stress stood at the 27th percentile. That gap says the plumbing measures look tighter than market pricing; it does not establish that a funding event has begun.

The fitted reserve-demand measure placed reserves $759B below its estimated kink as of 2026-08-12, with current reserves of $2,944.1B against a fitted kink of $3,703.1B. The fit’s R-squared was 0.616. This is a model estimate, not a revealed market threshold.

The nearest dated pressure point was 2026-08-20, when the reserve-path scenario included $266.0B of auction settlement. Seiche projected reserves of $2,935.2B, with a worst-case figure of $2,882.1B. The worst-case figures were $2,849.6B for 2026-08-31 and $2,831.1B for 2026-09-15. These are reserve-path scenarios, not observed outcomes.

Official-sector data add a separate footprint measure. Over 13 weeks through 2026-08-12, foreign officials cut Fed custody Treasuries by $90.2B while foreign reverse-repo parking rose $34.8B. The combined official footprint fell $55.5B to $2,954B. In the desk calculation, 62% of the custody decline left the official sector rather than rotating within it. The observation does not identify motive or establish a funding shock.

The selected historical episode has similarity of 0.561 and a lead of 30 days in the engine output. Those are construction point-in-time comparisons, not publication-vintage backtests or forecasts. Seiche’s analog work says analogs do not beat climatology on its sample; they are context, not a prediction tool.

## The strongest counter-case

The counter-case can defeat the tightening thesis: the observed price of overnight cash still says abundance. SOFR was 1.7bp below IORB as of 2026-08-12, rather than above it. Standing Repo Facility take-up was $0.00B as of 2026-08-14, so the backstop was not being tested in size.

Market stress was only at the 27th percentile of its own history as of 2026-08-13. The reported levels were 3.62% for SOFR, 3.63% for EFFR and 3.65% for IORB. Those observations are inconsistent with an overt overnight-funding shortage.

This is more than a caveat. The reserve-kink estimate can be wrong, or the calendar can be absorbed without disorder. If cash holders remain willing to lend, dealers retain capacity and market-clearing funding rates stay below IORB, the proposed transmission mechanism has not become a live constraint.

Confidence is low because a source or engine fault reduced the evidence set before interpretation. The reported fault concerned a BIS data request. That boundary weakens any claim that plumbing has decisively outrun price.

## A falsifiable next test

The test is not whether April 2025 repeats. After the 2026-08-20 settlement, observe whether the current plumbing signal gains confirmation in the overnight market and in facility use.

The tightening thesis gains support if SOFR no longer remains below IORB and Standing Repo Facility take-up rises from $0.00B. It also gains support if market stress rises from its 27th-percentile reading while the plumbing-versus-market gap closes.

The thesis is weakened if the settlement passes with SOFR still below IORB, Standing Repo Facility take-up still at $0.00B and market stress remaining subdued. The next dated check is the 2026-08-31 month-end window. Seiche’s published forecast was 2.2bp, within a band of -1.7bp to 6.7bp, with severity of 1 on a scale from 1 calm to 5 extreme. That is a low-severity published signal, not evidence of an imminent rupture.

## Follow the pressure chain

Start with the balance-sheet identity: Treasury General Account rebuilding and foreign reverse-repo absorption reduced reserves in the reported weekly ledger. Add auction settlement and other calendar pressure, producing dated demands on the reserve path. Then ask whether reserve scarcity changes funding behaviour; Seiche’s answer is model-based and uncertain, but the fitted distance from the kink is negative.

The next link is intermediation. The April 2025 replay shows the relevant condition: volatility, Treasury selling and relative-value deleveraging converge while dealers are asked to absorb the same shock. That is how a funding question can become an exit-capacity question.

The final link is observable market confirmation. Until overnight rates, facility use or market stress corroborate the plumbing read, the chain is incomplete. That is why this remains a desk brief rather than a full alarm.

## Sources, method, and limits

The underlying reading is in [Seiche’s daily dispatch](https://seiche.info/dispatches/2026-08-16-daily), with additional context in the [dispatch archive](https://seiche.info/dispatches/). Consult Seiche’s [data overview](https://seiche.info/data/overview.json), [series index](https://api.seiche.info/api/series/index.json) and [methodology](https://seiche.info/methodology).

The composite is a deterministic synthesis over published point-in-time engines. Confidence measures agreement across independent signals, not the probability of a market outcome. Source publication time is not collected uniformly, so event, desk-knowledge and publication clocks remain separate.

**Research boundary:** Seiche covers system dollar-funding capacity. [LiquiLens](https://liquilens.in/) covers institution and lender balance-sheet risk. [Undertow](https://liquilens-undertow.com/) covers market liquidity, crowding and exit capacity. Related historical readings are context-only and never enter the composite or a model feature.

The composite, reserve-path scenarios, fitted kink and historical comparison are Seiche derivations. Observed levels should not be confused with those derivations.

Research and market data, not investment advice.