From f97f5d2788de8067262ccd44e0747cbede6946d5 Mon Sep 17 00:00:00 2001
From: canvas-bot <canvas-bot@flow-master.ai>
Date: Mon, 15 Jun 2026 04:24:51 +0400
Subject: [PATCH] qa(idle-poll): assert zero background reqs in a 55s idle
 window
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Previous audit allowed up to 5 requests in a 30s window — generous,
hid real regressions. Tighten:

- 5s settle after each scene becomes interactive (covers the burst
  of mount-time fetches every scene legitimately fires)
- 55s observation window after settle, sized to fit between two
  consecutive 60s topbar badge polls
- MAX_REQ_PER_WINDOW = 0; any /api/* or /internal/* request during
  the window fails the assertion

This is the operator-experience contract: when you stop typing,
canvas stops fetching.
---
 qa/audit_idle_poll.mjs | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)

diff --git a/qa/audit_idle_poll.mjs b/qa/audit_idle_poll.mjs
index 31f58fd..ea9b537 100644
--- a/qa/audit_idle_poll.mjs
+++ b/qa/audit_idle_poll.mjs
@@ -1,8 +1,16 @@
 // Idle-poll audit. User explicitly complained "way too many GETs / non-stop".
-// For each scene: visit, settle, then watch network for 30s with no user input.
-// Pass if total requests in that 30s window is reasonable (<= 5).
-// The topbar identity-hydration + chat/queue badge polls at 60s, so a 30s
-// window should see at most 1 such fire.
+// For each scene: visit, settle for a generous burst window, then watch network
+// for 60s with no user input. Strict assertion: zero background /api/* or
+// /internal/* requests during that window.
+//
+// What counts as "background":
+//   - Anything fired more than SETTLE_MS after the scene becomes interactive
+//   - Excludes the 60s topbar badge poll fired exactly at minute boundaries —
+//     the window is sized to fit BETWEEN two such fires
+//
+// The topbar badge poll runs every 60s. A 55s observation window after a
+// 5s settle covers a full minute of operator inactivity without spanning
+// the next poll boundary, so the strict assertion "zero" is correct.
 import { chromium } from "playwright";
 
 const URL = "https://canvas.flow-master.ai/";
@@ -19,8 +27,9 @@ const SCENES = [
   { chip: /What is FlowMaster/, label: "explainer" },
 ];
 
-const IDLE_WINDOW_MS = 30_000;
-const MAX_REQ_PER_WINDOW = 5;
+const SETTLE_MS = 5_000;
+const IDLE_WINDOW_MS = 55_000;
+const MAX_REQ_PER_WINDOW = 0;
 
 const results = [];
 function record(name, ok, detail = "") {
@@ -49,7 +58,7 @@ for (const s of SCENES) {
 
   // Settle period — the scene's initial fetch burst should complete here.
   await p.waitForLoadState("networkidle").catch(() => {});
-  await p.waitForTimeout(1500);
+  await p.waitForTimeout(SETTLE_MS);
 
   // Now record every request fired for the next IDLE_WINDOW_MS.
   const requests = [];
@@ -65,9 +74,9 @@ for (const s of SCENES) {
 
   const ok = requests.length <= MAX_REQ_PER_WINDOW;
   const summary = requests.length
-    ? `${requests.length} reqs · ${requests.slice(0, 3).map((r) => `${r.method} ${r.url.slice(0, 50)}`).join(", ")}${requests.length > 3 ? "…" : ""}`
-    : "0 reqs · quiet";
-  record(`idle_${s.label}_quiet_at_30s`, ok, summary);
+    ? `${requests.length} background reqs in ${IDLE_WINDOW_MS / 1000}s · ${requests.slice(0, 3).map((r) => `${r.method} ${r.url.slice(0, 50)}`).join(", ")}${requests.length > 3 ? "…" : ""}`
+    : `0 background reqs in ${IDLE_WINDOW_MS / 1000}s`;
+  record(`idle_${s.label}_zero_background_traffic`, ok, summary);
 }
 
 await b.close();