diff --git a/.release-please-manifest.json b/.release-please-manifest.json
index a112e0abc4..effd9a1389 100644
--- a/.release-please-manifest.json
+++ b/.release-please-manifest.json
@@ -1,6 +1,6 @@
 {
   "packages/loopover-mcp": "3.14.1",
-  "packages/loopover-engine": "3.14.1",
+  "packages/loopover-engine": "3.15.0",
   "packages/loopover-miner": "3.14.1",
   "packages/loopover-ui-kit": "1.2.0"
 }
diff --git a/src/queue/processors.ts b/src/queue/processors.ts
index 70cc8736be..b35a438827 100644
--- a/src/queue/processors.ts
+++ b/src/queue/processors.ts
@@ -7348,6 +7348,39 @@ export function maybeAddRequiredAutoReviewSkipHold(
   return true;
 }
 
+/**
+ * #9015 — the REPUTATION skip's fail-closed hold, the exact sibling of the contributor-controlled skip
+ * above. A reputation downgrade (low signal, or the submissions>=8/merged<1 burst) suppresses AI review
+ * entirely; without this hold the PR then proceeds on deterministic checks alone — none of which read code
+ * semantics — and can auto-merge with ZERO defect detection. That inverts the feature's intent: a SUSPECTED
+ * abuser would receive LESS scrutiny than a trusted contributor. Where the repo requires blocking AI review,
+ * the skip must therefore hold for a human rather than silently pass. PURE (mutates the advisory it is
+ * given, like its sibling); the caller owns the reputation decision itself.
+ */
+export function maybeAddReputationSkipHold(
+  env: Env,
+  args: {
+    settings: RepositorySettings;
+    advisory: Pick<Awaited<ReturnType<typeof buildPullRequestAdvisory>>, "headSha" | "findings">;
+    repoFullName: string;
+    author: string | null;
+    confirmedContributor: boolean;
+    skipAiReview?: boolean | undefined;
+    reputationSkipped: boolean;
+  },
+): boolean {
+  if (!args.reputationSkipped || !shouldRequirePublicAiReviewForAdvisory(env, args)) return false;
+  args.advisory.findings.push({
+    code: "ai_review_inconclusive",
+    severity: "warning",
+    title: "Required AI review was skipped by a submitter-reputation downgrade",
+    detail:
+      "This repository requires blocking AI review, and the submitter's recent-submission signal downgraded this PR to deterministic-only checks. Those checks do not read code semantics, so the gate is held for human review instead of passing automatically.",
+    action: "Review this PR manually, or run AI review with a trusted override, before merging.",
+  });
+  return true;
+}
+
 /** Record a quiet auto-review skip (never a gate failure). Exported for unit tests. (#1954) */
 export async function auditPullRequestAutoReviewSkip(
   env: Env,
@@ -9926,6 +9959,19 @@ async function maybePublishPrPublicSurface(
       isReputationEnabled(env) && isConvergenceRepoAllowed(env, repoFullName)
         ? await shouldSkipAiForReputation(env, { project: repoFullName, submitter: author })
         : undefined;
+    // #9015: the reputation skip must FAIL CLOSED wherever blocking AI review is required — otherwise a
+    // suspected abuser's PR proceeds on deterministic checks alone (no code-semantics review at all) and can
+    // auto-merge, i.e. suspicion would BUY less scrutiny. Exact sibling of the contributor-controlled skip
+    // hold above; a no-op when the repo does not require blocking AI review, or when no skip fired.
+    maybeAddReputationSkipHold(env, {
+      settings,
+      advisory,
+      repoFullName,
+      author,
+      confirmedContributor,
+      skipAiReview: webhook.skipAiReview,
+      reputationSkipped: preComputedReputationSkip === true,
+    });
     // #one-shot-review-cadence: only even attempts the lookup when the review would otherwise be eligible to
     // run fresh this pass (mirrors how the frozen/paused branches below are similarly mutually exclusive) --
     // a PR that's blacklisted/frozen/already-skipped for another reason never shows AI content at all today,
diff --git a/src/review/submitter-reputation.ts b/src/review/submitter-reputation.ts
index 5ff0941789..7997def8b6 100644
--- a/src/review/submitter-reputation.ts
+++ b/src/review/submitter-reputation.ts
@@ -111,8 +111,13 @@ export function isMachinePacedCadence(cadence: SubmissionCadence): boolean {
 export async function getSubmitterCadence(env: Env, project: string, submitter: string | undefined): Promise<SubmissionCadence> {
   if (!submitter) return { count: 0, medianGapMs: null };
   try {
+    // #9015: reads the LIVE ledger. This query previously read `review_targets`, which stopped receiving
+    // writes at the 2026-06-22 self-host cutover — the cadence leg was silently inert (newest row frozen at
+    // the cutover date), so the machine-paced signal never fired for any submitter. `pull_requests` is the
+    // live per-PR table the same pipeline maintains; `created_at` is its ingest timestamp, which is what a
+    // SUBMISSION cadence is about.
     const result = await storage(env)
-      .prepare(`SELECT created_at AS createdAt FROM review_targets WHERE project = ? AND submitter = ? AND created_at >= datetime('now', ?) ORDER BY created_at DESC LIMIT ?`)
+      .prepare(`SELECT created_at AS createdAt FROM pull_requests WHERE repo_full_name = ? AND LOWER(author_login) = LOWER(?) AND created_at >= datetime('now', ?) ORDER BY created_at DESC LIMIT ?`)
       .bind(project, submitter, `-${CADENCE_WINDOW_HOURS} hours`, REPUTATION_WINDOW_ROW_CAP)
       .all<{ createdAt: string }>();
     const createdAts = (result?.results ?? []).map((r) => r.createdAt);
diff --git a/test/unit/auto-review-wiring.test.ts b/test/unit/auto-review-wiring.test.ts
index 42c9f4aba5..09e761fd62 100644
--- a/test/unit/auto-review-wiring.test.ts
+++ b/test/unit/auto-review-wiring.test.ts
@@ -2,6 +2,7 @@ import { describe, expect, it, vi } from "vitest";
 import {
   auditPullRequestAutoReviewSkip,
   maybeAddRequiredAutoReviewSkipHold,
+  maybeAddReputationSkipHold,
   resolveAutoReviewSkipForPullRequest,
   resolveReviewManifestForAiReview,
 } from "../../src/queue/processors";
@@ -405,6 +406,52 @@ describe("review.auto_review wiring (#1954)", () => {
 
     loadSpy.mockRestore();
   });
+  it("#9015: a reputation skip HOLDS where blocking AI review is required — suspicion must never buy less scrutiny", () => {
+    const blockingEnv = { AI_SUMMARIES_ENABLED: "true", AI_PUBLIC_COMMENTS_ENABLED: "true", AI: {} } as Env;
+    const settings = { gatePack: "oss-anti-slop", aiReviewMode: "block", aiReviewAllAuthors: false } as never;
+    const advisory = { headSha: "sha", findings: [] as unknown[] };
+    const added = maybeAddReputationSkipHold(blockingEnv, {
+      settings,
+      advisory: advisory as never,
+      repoFullName: "acme/widgets",
+      author: "burst-farmer",
+      confirmedContributor: false,
+      reputationSkipped: true,
+    });
+    expect(added).toBe(true);
+    expect(advisory.findings).toEqual([
+      expect.objectContaining({ code: "ai_review_inconclusive", severity: "warning", title: expect.stringContaining("submitter-reputation") }),
+    ]);
+    // No skip fired: nothing is added (the overwhelmingly common path).
+    const untouched = { headSha: "sha", findings: [] as unknown[] };
+    expect(
+      maybeAddReputationSkipHold(blockingEnv, {
+        settings,
+        advisory: untouched as never,
+        repoFullName: "acme/widgets",
+        author: "alice",
+        confirmedContributor: false,
+        reputationSkipped: false,
+      }),
+    ).toBe(false);
+    expect(untouched.findings).toEqual([]);
+    // AI review is OFF for this repo: nothing was expected to run, so a skip is not a suppression — silent,
+    // exactly like the contributor-controlled sibling (both share shouldRequirePublicAiReviewForAdvisory).
+    // In advisory mode the finding IS recorded and is non-blocking by nature there.
+    const reviewOff = { headSha: "sha", findings: [] as unknown[] };
+    expect(
+      maybeAddReputationSkipHold(blockingEnv, {
+        settings: { gatePack: "oss-anti-slop", aiReviewMode: "off", aiReviewAllAuthors: false } as never,
+        advisory: reviewOff as never,
+        repoFullName: "acme/widgets",
+        author: "burst-farmer",
+        confirmedContributor: false,
+        reputationSkipped: true,
+      }),
+    ).toBe(false);
+    expect(reviewOff.findings).toEqual([]);
+  });
+
   it("holds instead of quietly skipping when contributor-controlled metadata suppresses required AI review", () => {
     const advisory = { headSha: "sha", findings: [] };
     const added = maybeAddRequiredAutoReviewSkipHold(
diff --git a/test/unit/submitter-reputation.test.ts b/test/unit/submitter-reputation.test.ts
index a1b8e24fca..c5f28b8d2d 100644
--- a/test/unit/submitter-reputation.test.ts
+++ b/test/unit/submitter-reputation.test.ts
@@ -416,6 +416,22 @@ describe("getSubmitterCadence (D1, fail-safe) (#4514)", () => {
     expect(await getSubmitterCadence({} as Env, "p", undefined)).toEqual({ count: 0, medianGapMs: null });
   });
 
+  it("#9015: queries the LIVE pull_requests ledger, not the frozen review_targets table", async () => {
+    let sql = "";
+    const env = {
+      DB: {
+        prepare: (query: string) => {
+          sql = query;
+          return { bind: () => ({ all: async () => ({ results: [] }) }) };
+        },
+      },
+    } as unknown as Env;
+    await getSubmitterCadence(env, "acme/widgets", "farmer99");
+    expect(sql).toContain("FROM pull_requests");
+    expect(sql).not.toContain("review_targets");
+    expect(sql).toContain("author_login");
+  });
+
   it("derives cadence from the queried created_at timestamps", async () => {
     const t0 = new Date("2026-01-01T00:00:00.000Z").getTime();
     const env = makeCadenceEnv([t0, t0 + 5 * 60_000, t0 + 10 * 60_000, t0 + 15 * 60_000, t0 + 20 * 60_000].map((ms) => new Date(ms).toISOString()));