fix(ci): remove the unavailable self-hosted performance gate

Required CI waited unconditionally for a runner label backed by zero
registered runners, so every pull request remained queued forever. A
single non-ephemeral workstation is not a reliable merge prerequisite.

Keep deterministic benchmark smoke and safety ceilings in ordinary CI.
Keep the ten-round matched protocol as a documented, fail-closed local
command for deliberate performance work, but remove it from the required
workflow and its contract assertions.

Author: chh-ay

.changeset/fuzzy-mails-march.md

@@ -0,0 +1,4 @@
+---
+---
+
+Remove the unavailable self-hosted benchmark from required CI while retaining its local protocol.

.github/workflows/ci.yml

@@ -405,53 +405,6 @@ jobs:
             playwright-report/
           if-no-files-found: warn
 
-  controlled-performance:
-    name: Matched Zero-Regression Performance
-    needs: artifact-build
-    runs-on: sheetwrite-perf-i9-12900h-cachyos
-    timeout-minutes: 90
-    steps:
-      - name: Checkout
-        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-      - name: Setup Node
-        uses: actions/setup-node@249970729cb0ef3589644e2896645e5dc5ba9c38 # v6
-        with:
-          node-version: ${{ env.NODE_VERSION }}
-      - name: Setup Bun
-        uses: oven-sh/setup-bun@735343b667d3e6f658f44d0eca948eb6282f2b76 # v2.0.2
-        with:
-          bun-version: ${{ env.BUN_VERSION }}
-      - name: Install exact dependencies and benchmark browser
-        run: |
-          bun install --frozen-lockfile
-          bunx playwright install chromium
-      - name: Build controlled render harness
-        run: bun run --filter '@sheetwrite/bench' bench:render:prepare
-      - name: Capture repeated full matched samples
-        working-directory: bench
-        run: |
-          bun run src/render-driver.ts \
-            --rounds 10 \
-            --output results/render-fresh.json \
-            --markdown-output results/render-fresh.md
-      - name: Require matched zero-regression decision
-        working-directory: bench
-        run: |
-          bun run src/check.ts \
-            --baseline results/render-baseline.json \
-            --fresh results/render-fresh.json \
-            --power-mode balanced \
-            --concurrency 1
-      - name: Upload controlled raw samples
-        if: always()
-        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
-        with:
-          name: controlled-performance-raw-${{ github.sha }}
-          path: |
-            bench/results/render-fresh.json
-            bench/results/render-fresh.md
-          if-no-files-found: error
-          retention-days: 30
 
 
   required:
@@ -466,7 +419,6 @@ jobs:
       - delivery-size
       - docs-build
       - browser-smoke
-      - controlled-performance
     runs-on: ubuntu-latest
     timeout-minutes: 5
     steps:
@@ -480,7 +432,6 @@ jobs:
           SIZE: ${{ needs.delivery-size.result }}
           DOCS: ${{ needs.docs-build.result }}
           BROWSER: ${{ needs.browser-smoke.result }}
-          PERFORMANCE: ${{ needs.controlled-performance.result }}
           DOCS_REQUIRED: ${{ needs.preflight.outputs.docs_required }}
         run: |
           test "$PREFLIGHT" = success
@@ -489,7 +440,6 @@ jobs:
           test "$PACKED" = success
           test "$BUNDLERS" = success
           test "$SIZE" = success
-          test "$PERFORMANCE" = success
           if [ "$DOCS_REQUIRED" = "true" ]; then
             test "$DOCS" = success
             test "$BROWSER" = success

docs/src/content/docs/guides/performance-resources.md

@@ -4,9 +4,9 @@ description: "Freshness-gated benchmark and package-size evidence for Sheetwrite
 ---
 Every number on this page comes from a validated local protocol artifact captured on a clean tree; nothing is published from an unvalidated or protocol-mismatched artifact. Every expected cell carries either a validated timing or its recorded failure - a run that did not complete is shown as a failure, never converted into a timing.
 
-## Matched regression gate
+## Matched local regression check
 
-The release gate does not treat a competitor comparison or a smoke ceiling as regression evidence. On the controlled performance runner it captures ten fresh matched rounds, retains every raw sample, and compares the fresh artifact with the committed baseline. Any unapproved slowdown fails the required CI job.
+Timing comparisons run deliberately on a controlled local machine, not as a required CI job. Capture ten fresh matched rounds, retain every raw sample, and compare the fresh artifact with the committed baseline. Any unapproved slowdown fails the local command.
 
 ```sh verify title="Zero-regression benchmark"
 bun run --filter @sheetwrite/bench bench:render:prepare

scripts/docs.ts

@@ -1632,9 +1632,9 @@ export async function renderEvidencePage(sizeHistoryOverride?: SizeHistory): Pro
     ).trimEnd(),
     "Every number on this page comes from a validated local protocol artifact captured on a clean tree; nothing is published from an unvalidated or protocol-mismatched artifact. Every expected cell carries either a validated timing or its recorded failure - a run that did not complete is shown as a failure, never converted into a timing.",
     "",
-    "## Matched regression gate",
+    "## Matched local regression check",
     "",
-    "The release gate does not treat a competitor comparison or a smoke ceiling as regression evidence. On the controlled performance runner it captures ten fresh matched rounds, retains every raw sample, and compares the fresh artifact with the committed baseline. Any unapproved slowdown fails the required CI job.",
+    "Timing comparisons run deliberately on a controlled local machine, not as a required CI job. Capture ten fresh matched rounds, retain every raw sample, and compare the fresh artifact with the committed baseline. Any unapproved slowdown fails the local command.",
     "",
     '```sh verify title="Zero-regression benchmark"',
     "bun run --filter @sheetwrite/bench bench:render:prepare",

scripts/toolchain-contract.test.ts

@@ -158,26 +158,6 @@ describe("contributor and CI toolchain contract", () => {
     expect(commands).toContain("test:browser");
   });
 
-  it("requires a full matched comparison on the pinned performance runner", () => {
-    const jobs = parsedWorkflow.jobs ?? {};
-    const controlled = jobs["controlled-performance"];
-    expect(controlled?.["runs-on"]).toBe("sheetwrite-perf-i9-12900h-cachyos");
-    expect(controlled?.needs).toBe("artifact-build");
-    const commands = controlled?.steps?.flatMap((step) => (step.run ? [step.run] : [])).join("\n");
-    expect(commands).toContain("--rounds 10");
-    expect(commands).toContain("src/check.ts");
-    expect(commands).toContain("--power-mode balanced");
-    expect(commands).toContain("--concurrency 1");
-    expect(commands).not.toContain("--report-only");
-    expect(commands).not.toContain("bench:verify");
-    const upload = controlled?.steps?.find((step) =>
-      step.uses?.startsWith("actions/upload-artifact@"),
-    );
-    expect(upload?.if).toBe("always()");
-    expect(upload?.with?.path).toContain("render-fresh.json");
-    expect(JSON.stringify(jobs.required)).toContain("controlled-performance");
-  });
-
   it("gates the exact docs artifact without weakening Required CI", () => {
     const jobs = parsedWorkflow.jobs ?? {};
     const preflight = jobs.preflight;

scripts/workflow-contract.test.ts

@@ -179,10 +179,6 @@ describe("CI and release workflow contracts", () => {
       expect(commands(workflows().ci.jobs[jobName]!)).toContain("bun run browser:install:chromium");
     }
 
-    const controlledPerformance = workflows().ci.jobs["controlled-performance"]!;
-    expect(controlledPerformance["runs-on"]).toBe("sheetwrite-perf-i9-12900h-cachyos");
-    expect(commands(controlledPerformance)).toContain("bunx playwright install chromium");
-
     const projectByName = Object.fromEntries(
       (playwrightConfig.projects ?? []).map((project) => [project.name, project]),
     );