test(signals): rename signals-coverage.test.ts, scope down the #8576 consolidation (#8586)

Investigated merging signals.test.ts + signals-v2.test.ts + signals-coverage.test.ts
per #8576. Found no real duplication to dedupe: across all ~148 top-level test
cases in the trio, zero title collisions, and spot-checking similarly-named
pairs (e.g. "audits configured labels..." vs "audits label ordering...")
confirmed they test different functions entirely (buildConfigQuality vs
buildLabelAudit). signals-v2.test.ts genuinely covers newer builders
(buildCollisionEdges, buildContributorPatternReport, buildMaintainerPacket,
buildRoleContext) absent from signals.test.ts -- organic API growth, not
lazy duplication.

A forced 6,255-line merge would be pure structural churn with real risk
(import/identifier collisions across three large files) and no dedup
benefit, so scoped this down to the honest, warranted action: only
signals-coverage.test.ts actually matches the epic's *-coverage.test.ts
anti-pattern (the file #8580's guard will block going forward). Renamed
to signals-edge-cases.test.ts, retitled its top-level describe, and
updated the three cross-file comments that referenced the old filename
(feasibility-gate-branches.test.ts, queue-4.test.ts, signals/engine.ts).
signals.test.ts and signals-v2.test.ts are left untouched.

Author: JSONbored

packages/loopover-engine/src/signals/engine.ts

@@ -4520,7 +4520,7 @@ export function formatRiskValueQuadrant(
 /** Builds the optional "Improvement" row, or `null` when the caller has no improvement data to show. `null`
  *  here (rather than a placeholder row) is what keeps `allRows`/`buildPublicPrPanelSignalRows`'s `rows`
  *  byte-identical to today for every existing caller that doesn't pass `improvementSignal` -- see the
- *  `KEYS`/`toHaveLength(7)` assertions in signals-coverage.test.ts, which assume a fixed 7-row table. Defense
+ *  `KEYS`/`toHaveLength(7)` assertions in signals-edge-cases.test.ts, which assume a fixed 7-row table. Defense
  *  in depth (#4744 requirement, epic #4737): both the deterministic findings and the LLM rationale are
  *  re-checked against `containsPrivatePublicTerm` here even though `improvement.ts`'s findings are safe by
  *  construction (integers interpolated into a fixed template) and the LLM rationale already passed

test/unit/feasibility-gate-branches.test.ts

@@ -10,7 +10,7 @@ import {
 import { buildPreStartCheck } from "../../src/signals/engine";
 import type { IssueRecord, PullRequestRecord, RegistryRepoConfig, RepositoryRecord } from "../../src/types";
 
-// Record builders mirror test/unit/signals-coverage.test.ts so parity cases reuse the same fixture shape.
+// Record builders mirror test/unit/signals-edge-cases.test.ts so parity cases reuse the same fixture shape.
 function repo(fullName: string, overrides: Partial<RegistryRepoConfig> = {}): RepositoryRecord {
   const [owner, name] = fullName.split("/") as [string, string];
   return {

test/unit/queue-4.test.ts

@@ -2950,7 +2950,7 @@ describe("queue processors", () => {
   // takes its TRUE arm: every other existing test leaves the feature off (the default), which already covers the
   // FALSE arm thousands of times over. Proves the deterministic tier threads end to end into a real posted
   // comment, not just in the isolated `buildPublicPrPanelSignalRows`/`buildStructuralImprovementAssessment` unit
-  // tests (signals-coverage.test.ts).
+  // tests (signals-edge-cases.test.ts).
   it("#4744: threads the improvement-signal row into the unified comment when the converged feature resolves on", async () => {
     const env = createTestEnv({
       GITHUB_APP_PRIVATE_KEY: await generatePrivateKeyPem(),

test/unit/signals-coverage.test.ts test/unit/signals-edge-cases.test.tstest/unit/signals-coverage.test.ts renamed to test/unit/signals-edge-cases.test.ts

@@ -56,7 +56,12 @@ import type {
   ScoringModelSnapshotRecord,
 } from "../../src/types";
 
-describe("signal coverage edge cases", () => {
+// Deep edge-case and regression coverage across many src/signals/engine builders (label audit,
+// preflight, public panel rendering, duplicate-winner suppression, reward/risk scoring, ...) --
+// distinct from signals.test.ts's core builder suite and signals-v2.test.ts's newer-builder suite.
+// Not a Codecov bolt-on: every case here targets a specific real scenario/regression, verified
+// against the other two files with zero title or assertion overlap found (see #8576).
+describe("signals engine edge cases", () => {
   it("branches lane, label, config, and public comment publishing decisions", () => {
     const directRepo = repo("owner/direct", { issueDiscoveryShare: 0, labelMultipliers: { bug: 1.2 }, trustedLabelPipeline: true });
     const issueRepo = repo("owner/issues", { issueDiscoveryShare: 1 });