ci(nightly-bench): runs-on perf-solo (re-cut of #6409 on current main, Refs #6407)

.github/workflows/nightly-bench.yml

@@ -60,7 +60,7 @@ jobs:
     concurrency:
       group: perf-intel
       cancel-in-progress: false
-    runs-on: [self-hosted, X64, Linux, clean-room, intel]   # perf: the SAME box as the recorded baselines — clean-room now spans intel + yoga + gx10 (#3100)
+    runs-on: [self-hosted, X64, Linux, intel, perf-solo]   # REV 2.0 step 6: same selector as beat-speed-nightly.yml:68; perf-solo = one runner, so two timings can never overlap
     timeout-minutes: 45
     steps:
       - uses: actions/checkout@v7

docs/roadmaps/entries/PMAT-6407.yaml

@@ -0,0 +1,19 @@
+- id: PMAT-6407
+  github_issue: 6407
+  item_type: task
+  title: REV 2.0 step 6 nightly-bench runs-on perf-solo
+  status: in_progress
+  priority: high
+  assigned_to: null
+  created: 2026-10-26T11:03:41Z
+  updated: 2026-10-26T11:03:41Z
+  spec: null
+  acceptance_criteria:
+  - nightly-bench measure job selects [self-hosted, X64, Linux, intel, perf-solo] and lands on intel-clean-room-16
+  phases: []
+  subtasks: []
+  estimated_effort: null
+  labels:
+  - kind:code
+  notes: null
+  epic: 6390

docs/roadmaps/roadmap.yaml

@@ -26434,6 +26434,25 @@ Not done, pending the ruling: moving the 31 fixture files to an obviously synthe
   - kind:code
   notes: null
   epic: 6390
+- id: PMAT-6407
+  github_issue: 6407
+  item_type: task
+  title: REV 2.0 step 6 nightly-bench runs-on perf-solo
+  status: in_progress
+  priority: high
+  assigned_to: null
+  created: 2026-10-26T11:03:41Z
+  updated: 2026-10-26T11:03:41Z
+  spec: null
+  acceptance_criteria:
+  - nightly-bench measure job selects [self-hosted, X64, Linux, intel, perf-solo] and lands on intel-clean-room-16
+  phases: []
+  subtasks: []
+  estimated_effort: null
+  labels:
+  - kind:code
+  notes: null
+  epic: 6390
 - id: PMAT-6410
   github_issue: 6410
   item_type: task