From bb33626f3806fbde5ecdfb3e65e99669c99eae71 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 30 Sep 2026 08:16:19 +0300
Subject: [PATCH] ci(perf): the interaction first-frame ceiling sits above the
 1.78 level (#692)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The full workflow's 7-sample median of firstStableRenderMs for
large-house-interaction-v1 was about 2790 ms across the 1.77 line and
about 2920 ms at v1.78.0 (7d4d75bd: 2925.0; the neighbouring run of the
same SHA read 3144.8). The 3000 ms ceiling sat 2.7 % above the level and
failed on runner noise; #689's 3-sample smoke read 3002.4.

- hardMaxMs 3000 → 3400 in the full profile and its smoke twin (one
  number, #473 AC4): +16 % over the 1.78 level, +8 % over the worst run.
  The base-relative ratio and noise allowance are unchanged.
- README: the series and the reasoning; the test pins the number and the
  series points.

Issue: #692
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
---
 demo/performance/README.md                     | 11 +++++++++++
 .../performance/budgets-interaction-smoke.json |  2 +-
 .../budgets-large-house-interaction.json       |  2 +-
 test/performance-budget.test.mjs               | 18 ++++++++++++++++++
 4 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/demo/performance/README.md b/demo/performance/README.md
index 0b93832a..c2a0f271 100644
--- a/demo/performance/README.md
+++ b/demo/performance/README.md
@@ -99,6 +99,17 @@ 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.
 
+Its `firstStableRenderMs` ceiling is 3400 ms (#692). The full workflow's
+7-sample median on the hosted runner was about 2790 ms across the 1.77 line
+(2769.0 on 09-21, 2838.2 on 09-23, 2777.3 and 2801.7 on `c392ad3a`) and about
+2920 ms at v1.78.0 (`7d4d75bd`: 2925.0, 2925.2 as the base). Two runs of that
+one SHA differed by 7.5 % (3144.8 vs 2925.0), so the former 3000 ms ceiling
+sat 2.7 % above the level and failed on noise — the 3-sample smoke of #689
+read 3002.4 ms. 3400 ms is +16 % over the 1.78 level and +8 % over the worst
+run in the series; the base-relative comparison of the full workflow is
+unchanged and remains the detector for smaller regressions. The +4.5 % step
+between 1.77 and 1.78.0 is not attributed here.
+
 The isometric `spaceSwitchMs` ceiling is 2200 ms, shared by the smoke, the full
 isometric profile and its Stage 3 dense twin (#675). The metric is one cold
 floor switch that builds the second floor's 2.5D geometry, so it grows with
diff --git a/demo/performance/budgets-interaction-smoke.json b/demo/performance/budgets-interaction-smoke.json
index cdcc7aff..0ae8e069 100644
--- a/demo/performance/budgets-interaction-smoke.json
+++ b/demo/performance/budgets-interaction-smoke.json
@@ -9,7 +9,7 @@
     },
     "firstStableRenderMs": {
       "stat": "median",
-      "hardMaxMs": 3000
+      "hardMaxMs": 3400
     },
     "interactionSeriesMs": {
       "stat": "median",
diff --git a/demo/performance/budgets-large-house-interaction.json b/demo/performance/budgets-large-house-interaction.json
index 9b85f748..929f985d 100644
--- a/demo/performance/budgets-large-house-interaction.json
+++ b/demo/performance/budgets-large-house-interaction.json
@@ -13,7 +13,7 @@
       "stat": "median",
       "maxRegressionRatio": 0.3,
       "noiseAllowanceMs": 250,
-      "hardMaxMs": 3000
+      "hardMaxMs": 3400
     },
     "interactionSeriesMs": {
       "stat": "median",
diff --git a/test/performance-budget.test.mjs b/test/performance-budget.test.mjs
index d44b6b50..4229e823 100644
--- a/test/performance-budget.test.mjs
+++ b/test/performance-budget.test.mjs
@@ -251,6 +251,24 @@ for (const [smokeName, fullName] of [
   });
 }
 
+test('#692 первый стабильный кадр interaction: потолок над уровнем 1.78 с запасом на шум раннера', () => {
+  const smoke = readBudget('budgets-interaction-smoke.json');
+  const full = readBudget('budgets-large-house-interaction.json');
+  const ceiling = full.timings.firstStableRenderMs.hardMaxMs;
+  assert.equal(ceiling, 3400);
+  assert.equal(smoke.timings.firstStableRenderMs.hardMaxMs, ceiling, 'смок = полный профиль (#473 AC4)');
+  // Медианы полного профиля (7 образцов) из performance.yml: линия 1.77 и v1.78.0.
+  const level177 = [2769.0, 2838.2, 2777.3, 2801.7];
+  const level178 = [2925.0, 2925.2];
+  const worst = 3144.8; // тот же SHA v1.78.0, соседний прогон — шум раннера до 7,5 %
+  for (const median of [...level177, ...level178, worst]) assert.ok(median < ceiling, `${median} краснеет на ${ceiling}`);
+  assert.ok(ceiling >= Math.max(...level178) * 1.15, 'запас над уровнем 1.78 не меньше 15 %');
+  assert.ok(ceiling <= Math.max(...level178) * 1.2, 'и не больше 20 %: потолок остаётся гардом');
+  // Относительное сравнение полного профиля не ослаблено.
+  assert.equal(full.timings.firstStableRenderMs.maxRegressionRatio, 0.3);
+  assert.equal(full.timings.firstStableRenderMs.noiseAllowanceMs, 250);
+});
+
 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');