# A4S-1 — the A4 days-to-cover vacuity gate has been built, merged, pre-registered and returns-free since 2026-08-01, and nobody has run it: the memo assigns it to "the operator", which is not an actor this system has

status: **dismissed as a queue row** · raised panel run 42 (2026-10-21;
surfaced by the frontier lens as a WATCH fact, verified and argued by the
chair) · killed at the chair's own adversarial pass on CAP + NO-DEADLINE
grounds, **not** on merit · class: registration hygiene (G5) — a pinned test
with no assigned actor · effort would have been ~0 pd to run, ~0.1 pd to
record the verdict

## What is true (verified at `9e28a45`, by execution where possible)

`research/2026-08-01_a4_dtc_battery_memo.md` §5 pins a two-step ordering it
calls non-negotiable. **Step 1** is a vacuity count that must run before any
return is seen, and the memo says of it, verbatim (`RESEARCH.md:852-855`):

> *"**(3) The tool:** `thales dtc-vacuity-count` (`backtest/dtc_vacuity.py`)
> computes the count WITHOUT touching returns (selections ∩ lagged-DTC sets
> only), lockstep-tested against the engine's actual screen; **the operator
> runs it post-merge — its author did not run it.**"*

| check | result |
|---|---|
| the tool exists and is wired to the CLI | **CONFIRMED-BY-EXECUTION** — `thales --help` lists `dtc-vacuity-count  A4 days-to-cover vacuity count — returns-free;` |
| a verdict was ever recorded | **NO.** `grep -rln "dtc.vacuity\|dtc_vacuity\|vacuity.count" research/ RESEARCH.md results/` returns only the two 2026-08-01 memos, `RESEARCH.md`'s own registration, and RAD-1 (which cites it as a *template*). No verdict memo, nothing in `results/diagnostics/` |
| a trial-registry row exists | **NO.** 0 of 415 rows in `results/research_registry.jsonl` match `dtc` or `a4` |
| prior art in the queue | **NONE.** No row in `open/`, `approved/`, `built/` or `dismissed/`; 0 hits in the frozen `research/RESEARCH_QUEUE.md` |

**81 days.** The gate is returns-free by construction (selections and set
intersections only), its thresholds are pinned in the merged memo, and its
expected modal outcome is recorded at freeze as *"vacuous-or-inconclusive — a
legitimate cheap closure."* `< 1.0` name-change per covered quarter ⇒ VACUOUS,
the stream stays capture-only, **the battery never runs**, and the registration
closes. `>= 1.0` authorizes Step 2 and nothing more.

**It is the blocker for two deferred directions.** A4's battery cannot start
without it, and B3 (the FTD probe) attaches to that battery as a reported-only
second trial — so B3's host does not exist until this is pulled.

**Why nobody pulled it:** the memo names "the operator". Under the 2026-10-07
delegation the weekday loop is panel / triage / implementer / daily audit /
steward, and none of them is that. The steward may not run it (`ops/STEWARD.md`
§0), no routine container has `data/raw` (the momentum composite needs it, so
**NOT-RUN here, and not guessable**), and the implementer builds only from
`approved/`. This is CLK-1's defect — *a criterion that fires into a passive
verb* — applied to a test rather than a clock.

## Why it is dismissed as a ROW

1. **It has no deadline at all.** The memo's own horizon is "earliest ship
   post-December." Triage's ranking rule is *"something cheap that expires
   beats something valuable that will wait"* (`ops/QUEUE_TRIAGE.md:150-152`),
   and this waits. It has waited 81 days with no cost beyond the waiting.
2. **`open/` was at 12 of 12** and this run already spends the one slot it
   could justify on DLV-1, which bites this Friday. Opening a second
   no-deadline row would have forced an eviction, and the pre-declared
   eviction's rationale is itself falsified (run 41 on DGX-1).
3. **A row would not produce the run.** The row's entire content is "someone
   with `data/raw` types one command." The queue's builder has not run since
   2026-10-07, and two owner-approved rows (FRZ-1, OSR-1) are already waiting
   on it. Adding a third item that needs an actor the queue cannot supply
   makes the queue longer, not the question closer.
4. **The honest cadence disclosure.** This was surfaced by the frontier lens,
   so filing it as an open row would nominally have reset QUEUE_TRIAGE counter
   (c) — the criterion that FIRED on 2026-10-19 and drops this panel from daily
   to weekly. **It does not reset it and this run does not claim that it does.**
   A threshold that has fired is not un-fired by a later item; relitigating it
   is nobody's (`ops/QUEUE_TRIAGE.md:132-133`). Recorded explicitly so no
   future reader can mistake this dismissal for a cadence defence, and so the
   panel is not seen scoring its own exam.

**What this dismissal does instead, at zero cost:** it is the dated, greppable,
in-repo record that the gate is unrun and why — which is most of what a row
would have delivered. `dismissed/` is swept by triage on a 14-day cadence and
read by every panel, so this reaches a reader; a line in a session report does
not.

## Reopen conditions (any one)

1. **An interactive or owner-side session with `data/raw` is running anyway.**
   Then this is not a queue item at all — it is one command and a recorded
   verdict, and it should simply be done: `thales dtc-vacuity-count`, append the
   statistic and the VACUOUS/AUTHORIZED verdict to `RESEARCH.md` and the trial
   registry. **Do not re-open a row to schedule a one-command gate.**
2. **December passes and A4 is still named as a live direction** anywhere
   (`NORTHSTAR.md` §6 carries "A4 DTC battery earliest ship — post-December").
   At that point the gate is on the critical path of a dated clock and earns a
   slot on its own.
3. **B3 is proposed, revived, or asked after.** Its host cannot exist until
   Step 1 runs, so any serious interest in B3 reopens this first, and B3 must
   not be worked before it.
4. **The memo's ordering is at risk of being violated** — i.e. anyone proposes
   computing any return, equity curve or performance statistic for a
   DTC-screened variant. Step 1 is the gate that makes Step 2 legitimate; a
   Step 2 before it is a C0 breach, and this file is the standing tripwire.

**Explicitly NOT a reopen condition:** the passage of time alone. 81 days cost
nothing and 181 will cost nothing. Only conditions 1–4 make it matter.