Where Did the Trace Break? An Attempt-Bound, Fail-Closed Measurement Gate for Noetis

Can a boundary-tapped, attempt-bound, atomic measurement gate reconstruct every Noetis provider attempt and refuse audit-qualified release when the trace is incomplete, at fixed model compute?

Proposes the Attempt-Bound Measurement Gate (ABMG). A static audit of the current collector found 10 of 31 requirements present, 9 partial and 12 absent. D21-A is an executed deterministic development diagnostic: 34 of 34 named checks passed. Zero language-model calls, zero provider network calls, zero production-database calls. These are constructed conformance findings, not estimates of production incident rates. E21-B and H21-1 to H21-6 are proposed and untested. No model benchmark, production integration test, security audit, Noetis code change, or deployment.

• Under the current three-write best-effort persistence model, seven of eight possible write states are incomplete yet do not block release; ABMG audit-qualifies none of those seven. Logical state-space result, not a measured production rate.
• Ten envelope-only mutations collide under the current content-only fingerprint, 24 equal-byte different-version source pairs and 24 distinct provider-response pairs collapse under the current records, and five bad stream fixtures are falsely accepted by a development model of the current parser; the ABMG projection separates or rejects all of them. Constructed fixtures, not production results.

• E21-A Deterministic attempt-bound measurement gate diagnostic (D21-A) Executed (diagnostic)
• E21-B Shadow ABMG evaluation on real provider traffic at fixed model compute Proposed

• H21-1 The boundary-tapped ABMG collector reconstructs at least 99.5 percent of provider attempts exactly and improves reconstructability by at least 20 percentage points over the current collector on the same traffic.
• H21-2 Under injected single-write, process-crash, and retry-race failures, false audit_qualified_released is at most 0.1 percent with a two-sided 95 percent upper confidence bound at or below 0.5 percent.
• H21-3 Under truncated streams, malformed events, and HTTP-200-then-error cases, false-complete classification is at most 0.1 percent and recall for valid complete streams is at least 99.5 percent.
• H21-4 With model calls held identical, A1 instrumentation adds at most 5 ms median and 20 ms p95 synchronous latency, at most 15 KB median content-free storage per provider call, and no statistically or practically meaningful answer-quality regression.
• H21-5 Reviewers given ABMG traces reconstruct attempt lineage and terminal status at least 30 percent faster, with at least 95 percent correctness, than reviewers given current traces.
• H21-6 The boundary tap outperforms an otherwise equivalent post-hoc collector under envelope mutation and fallback.

Proposed measurement contract only. No Noetis code, inference, routing, billing, collector, or release behavior was changed.

• Does ABMG capture at least 99.5 percent of live attempts exactly?
• Does synchronous gate overhead meet budget at fixed model compute?
• Does failure labeling improve reviewer reconstruction, and does the privacy design withstand review?

Builds on Round 20, Round 19, Round 10, Round 9. Continued in Round 23, Round 22.