perf(desktop): baseline multitab + render-churn, leave idle-cost report-only

Captures medians of 5 runs for multitab and render-churn so tonight's
wins can't silently regress.

idle-cost is deliberately NOT gated. Its render attribution and idle
commit rate are trustworthy and are what the scenario exists for, but the
drag fps it reports (~0.6fps, p95 814ms) contradicts a direct
single-clock probe of the same gesture on the same build (57fps). I ruled
out sash selection, tile setup, render-counter residue, and a 20s soak,
and could not explain the gap — so the metric ships as a report, not a
gate. Gating CI on a number I can't defend would either fire on a phantom
or mask a real stall.

tier: 'report' is outside GATED ('ci','cold'), so the scenario still runs
and prints but neither compares nor writes a baseline.

apps/desktop/scripts/perf/README.md

@@ -52,6 +52,7 @@ directly via `window.__PERF_DRIVE__`, so no LLM credits are spent.
 | `keystroke` | ci | composer keystroke → paint latency | measure-latency, profile-typing, leak-typing |
 | `transcript` | ci | large-transcript mount + paint cost | (new) |
 | `render-churn` | ci | per-component render attribution + store churn while N tabs stream | (new) |
+| `idle-cost` | report | busy-but-silent tiles: idle commit rate, + fps while resizing / typing | (new) |
 | `cold-start` | cold | launch → CDP → driver → first paint (fresh spawn/run) | (new) |
 | `first-token` | backend | Enter → first assistant token painted (TTFT) | (new) |
 | `submit` | backend | Enter → cleared → user msg painted, scroll jump | measure-submit, measure-jump |

apps/desktop/scripts/perf/baseline.json

@@ -1,9 +1,9 @@
 {
   "_meta": {
-    "note": "Median of 5 runs, darwin-arm64, `--spawn --prod` (PRODUCTION minified renderer, real boot — no fake-boot). Representative shipped numbers, not dev-inflated. cold-start reuses one profile so the V8 code cache is WARM (what users get after first launch, ~1.0s); a fresh-profile first launch is ~+400ms (measure with `--cold-fresh`). Marks are process-spawn wall clock (spawn_to_*) or renderer nav-relative (dom_*). Re-baseline per device with `--update-baseline`; tolerances loose for cross-machine/disk variance.",
+    "note": "Median of 5 runs, darwin-arm64, `--spawn --prod` (PRODUCTION minified renderer, real boot \u2014 no fake-boot). Representative shipped numbers, not dev-inflated. cold-start reuses one profile so the V8 code cache is WARM (what users get after first launch, ~1.0s); a fresh-profile first launch is ~+400ms (measure with `--cold-fresh`). Marks are process-spawn wall clock (spawn_to_*) or renderer nav-relative (dom_*). Re-baseline per device with `--update-baseline`; tolerances loose for cross-machine/disk variance.",
     "platform": "darwin-arm64",
     "node": "v24.11.0",
-    "updated": "2026-07-19T23:16:01.227Z"
+    "updated": "2026-07-27T00:30:11.290Z"
   },
   "scenarios": {
     "stream": {
@@ -55,6 +55,25 @@
         "dom_content_loaded_ms": 574,
         "nav_to_read_ms": 721
       }
+    },
+    "multitab": {
+      "metrics": {
+        "longtasks_n": 0,
+        "longtask_max_ms": 0,
+        "frame_p95_ms": 29.1,
+        "frame_p99_ms": 36.2,
+        "slow_frames_33": 9
+      }
+    },
+    "render-churn": {
+      "metrics": {
+        "sidebar_renders": 0,
+        "sidebar_wasted": 0,
+        "wasted_renders": 1704,
+        "total_renders": 8221,
+        "commits": 1352,
+        "wasted_notifies": 0
+      }
     }
   }
 }

apps/desktop/scripts/perf/scenarios/idle-cost.mjs

@@ -202,7 +202,13 @@ const round = (n, places = 1) => Math.round(n * 10 ** places) / 10 ** places
 
 export default {
   name: 'idle-cost',
-  tier: 'ci',
+  // NOT 'ci': the drag fps this reports (~0.6fps, p95 814ms) contradicts a
+  // direct single-clock probe of the same gesture on the same build (57fps),
+  // and I could not reconcile the two — ruled out sash selection, tile setup,
+  // counter residue, and a 20s soak. Its RENDER attribution and idle commit
+  // rate are trustworthy and are what this scenario is for; the interaction
+  // fps is reported for investigation, not gated on, until that is explained.
+  tier: 'report',
   description: 'Busy-but-silent tiles: idle commit rate, and fps while resizing / typing.',
   async run(cdp, opts = {}) {
     const tiles = Number(opts.tiles ?? 5)