fix(test): stabilize default timeouts and perf budget

CHANGELOG.md

@@ -27,6 +27,7 @@ This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.htm
 - Tradeoff: dropped Node 18 from required support because `commander@14` (runtime dependency) requires Node `>=20`; preserving `18/20/22` would create false support claims and non-deterministic install/runtime behavior.
 - `drift doctor --json` and `drift guard --json` now emit schema metadata (`$schema`, `toolVersion`) and are covered by v1 schema contract tests (`schemas/drift-doctor.v1.json`, `schemas/drift-guard.v1.json`).
 - Added performance regression gate (`npm run check:perf-budget`) with versioned budgets in `benchmarks/perf-budget.v1.json`, benchmark memory/runtime contract checks, and CI artifact upload under `.drift-perf/` (gated on Node 20 to reduce matrix flakiness).
+- Stabilized local/CI test reliability by setting global Vitest default timeouts (15s) and tuning the scan runtime perf budget threshold for lower false-positive noise.
 
 ---
 

benchmarks/perf-budget.v1.json

@@ -12,7 +12,7 @@
   },
   "tasks": {
     "scan": {
-      "maxMedianMs": 3500,
+      "maxMedianMs": 3800,
       "maxRssMb": 700
     },
     "review": {

vitest.config.ts

@@ -6,6 +6,8 @@ export default defineConfig({
     globals: true,
     environment: 'node',
     include: ['tests/**/*.test.ts'],
+    testTimeout: 15000,
+    hookTimeout: 15000,
     coverage: {
       provider: 'v8',
       include: ['src/**/*.ts'],