diff --git a/demo/performance/README.md b/demo/performance/README.md
index 4ad4f289..ea41b91c 100644
--- a/demo/performance/README.md
+++ b/demo/performance/README.md
@@ -92,6 +92,13 @@ the full profiles. The set of profiles is part of the `performance_smoke`
 reuse key, so a glow-only success never stands in for a run that needed the
 isometric profile.
 
+The interaction profile keeps strict per-series, Long Task, heap, cache and
+structural limits. Its aggregate `interactionSeriesMs` ceiling is 3300 ms:
+enough headroom for the observed hosted-runner baseline (up to 3122.6 ms), but
+still below the known failed pre-optimization result (3501.5 ms). This aggregate
+is a catastrophic guard; the full workflow's base-relative comparison remains
+the detector for smaller regressions.
+
 The dedicated `performance.yml` workflow is the full comparison. It runs on
 every `main` promotion, weekly and on manual dispatch for an important beta or
 performance-sensitive change. It checks out the candidate and its base SHA,
diff --git a/demo/performance/budgets-interaction-smoke.json b/demo/performance/budgets-interaction-smoke.json
index 38c03270..cdcc7aff 100644
--- a/demo/performance/budgets-interaction-smoke.json
+++ b/demo/performance/budgets-interaction-smoke.json
@@ -13,7 +13,7 @@
     },
     "interactionSeriesMs": {
       "stat": "median",
-      "hardMaxMs": 3000
+      "hardMaxMs": 3300
     },
     "hoverSeriesMs": {
       "stat": "median",
diff --git a/demo/performance/budgets-large-house-interaction.json b/demo/performance/budgets-large-house-interaction.json
index 8ccfcd10..9b85f748 100644
--- a/demo/performance/budgets-large-house-interaction.json
+++ b/demo/performance/budgets-large-house-interaction.json
@@ -19,7 +19,7 @@
       "stat": "median",
       "maxRegressionRatio": 0.5,
       "noiseAllowanceMs": 150,
-      "hardMaxMs": 3000
+      "hardMaxMs": 3300
     },
     "hoverSeriesMs": {
       "stat": "median",
diff --git a/test/performance-budget.test.mjs b/test/performance-budget.test.mjs
index bbedf049..a66b545a 100644
--- a/test/performance-budget.test.mjs
+++ b/test/performance-budget.test.mjs
@@ -245,3 +245,14 @@ for (const [smokeName, fullName] of [
       'первый кадр 9 870 мс обязан краснеть');
   });
 }
+
+test('interaction aggregate keeps hosted-runner headroom without weakening component gates (#483)', () => {
+  const smoke = readBudget('budgets-interaction-smoke.json');
+  const full = readBudget('budgets-large-house-interaction.json');
+  assert.equal(smoke.timings.interactionSeriesMs.hardMaxMs, 3300);
+  assert.equal(full.timings.interactionSeriesMs.hardMaxMs, 3300);
+  assert.equal(smoke.timings.hoverSeriesMs.hardMaxMs, 500);
+  assert.equal(smoke.timings.panSeriesMs.hardMaxMs, 500);
+  assert.equal(smoke.timings.cameraSeriesMs.hardMaxMs, 500);
+  assert.equal(smoke.timings.editorSeriesMs.hardMaxMs, 750);
+});