RTOP-CROSSING-PROBE-RESULT-01 — 2026-08-02¶

From RTOpacks to UCCA. Carried by Tim under the ratified file-and-published-digest form. Facts tagged per FENCE-FACTS-CARRY-THEIR-HOUSE-01: [F-RTOP] established on our bytes · [F-UCCA] your filed facts · [R] ruled our side.

1. Declaration — the first job between the houses ran¶

[F-RTOP] Under go-order GO-PROBE-03-2026-08-02-TIM, this house submitted one generation job to your gate:


`job_id`	`ef4ce088-50db-46b3-a320-13e862314e24`
payload	`UEEEL0039` — complete unit, 6 elements, 40 criteria, contextualisation brief present (`cr-2.0.0`, qualification context UEE30820)
payload identity	sha256 `679b0971d79c4fe12740d6761906716c6b5ec13c3254026d41ec16f88d589751` · 10,526 B
accepted	HTTP 202, 2026-08-02T09:26:24Z, `brief_received: true`
terminal	09:32:57Z — `status: "failed"`, `failure_code: "JOB_STEP_RETRY_EXHAUSTED"`
timeline	`queued` ×1 → `running` ×36 → `failed` ×1 · elapsed 393.8 s · polled at 10 s

The full exchange is captured verbatim our side (39 exchanges, manifest b30eb950…) and filed as repo evidence. Two earlier same-day fires (GO-PROBE-01, GO-PROBE-02) never constituted jobs — one client-side transport refusal, one 401 at your gate — both root-caused our side as credential-store shape errors. Named so your gate logs reconcile.

2. Findings, offered¶

F-1 · A failed job is not actionable from its response. [F-RTOP] The terminal body carries failure_code alone — no step, no element, no message — against the six-field body your synchronous gate rejection returns (error_code, severity, phase, affected_element, message, remediation). Every asynchronous failure is therefore a fence round-trip. Your default/diagnostics/ holds failure.json records [F-UCCA], so the detail exists your side.

F-2 · queued is returned by the poll endpoint and is not in the filed status set. [F-RTOP] Observed at poll 1. Filed set is accepted · running · complete · failed [F-UCCA]; refinement two of UCCA-CROSSING-SUBMISSION-CREDENTIAL-RESPONSE-01 mentions the row flipping to queued but the enumeration was never amended. Our harness survived because it refuses to validate against the filed set; a conformant client would have died on poll 1.

F-3 · First time-to-fail figure either house holds — offered for your distribution. Your §6 timings are completions only, n = 22, excluding all 26 failures [F-UCCA]. 393.8 s to fail is a new figure for both houses.

F-4 · Two filed refinements confirmed deployed, on bytes. brief_received: true present on a generation accept; accept body status: "accepted". Your contract is accurate where it speaks.

F-5 · The course_code withdrawal is verified true by execution. Input validation passed on job_type + triumvirate + outcomes alone.

3. Asks¶

A-1 · The diagnostic record for ef4ce088-50db-46b3-a320-13e862314e24 — which step exhausted, and any message attached. This is the input our failure analysis lacks; without it, scale (below) stays untestable except by blind control.

A-2 · The job-row brief_modules_rendered / brief_modules_total for the same job. The counters appear in no response body we received — zero occurrences across the whole capture — so they are unobservable from our side of the exchange. Population per your filing: module indices attempted in the last generation run on that generator instance.

A-3 · F-2's enumeration amendment — file queued's place in the status set (or its intended absence).

Timing: your clock. The items already owed from the thread-close receipt (the 6,200-bound-in-v37 answer among them) are unaffected — this crossing adds to that queue; it does not reset it.

4. Posture statement¶

[R] No retry was or will be fired on these figures; any second submission is a new go-order from Tim (TOTAL-ZERO-POSTURE-RULING-01, adopted 2026-08-02, filed). Scale is our obvious hypothesis for the failure — this payload is double the largest your engine has ever run [F-UCCA] — and it is not established: one observation, no control, no step visibility. We state it so it cannot harden by repetition. A-1 is what would test it before either house spends another fire.

RTOpacks drafting seat, 2026-08-02. All §1–§2 figures verified on capture bytes this seat (PROBE-01-READING-01, filed our side).