fix(qa): serialize runtime parity cells

extensions/qa-lab/src/runtime-parity.test.ts

@@ -138,6 +138,25 @@ describe("runtime parity", () => {
     expect(result.drift).toBe("none");
   });
 
+  it("runs runtime cells serially so shared QA state cannot cross-contaminate", async () => {
+    const events: string[] = [];
+    const result = await runRuntimeParityScenario({
+      scenarioId: "serial",
+      runCell: async (runtime) => {
+        events.push(`start:${runtime}`);
+        await Promise.resolve();
+        events.push(`finish:${runtime}`);
+        return {
+          scenarioStatus: "pass",
+          cell: makeCell(runtime),
+        };
+      },
+    });
+
+    expect(result.drift).toBe("none");
+    expect(events).toEqual(["start:pi", "finish:pi", "start:codex", "finish:codex"]);
+  });
+
   it("classifies final-text-only differences as text-only", async () => {
     const result = await runRuntimeParityScenario({
       scenarioId: "text-only",

extensions/qa-lab/src/runtime-parity.ts

@@ -910,7 +910,8 @@ export async function runRuntimeParityScenario(params: {
   scenarioId: string;
   runCell: (runtime: RuntimeId) => Promise<RuntimeParityScenarioExecution>;
 }): Promise<RuntimeParityResult> {
-  const [pi, codex] = await Promise.all([params.runCell("pi"), params.runCell("codex")]);
+  const pi = await params.runCell("pi");
+  const codex = await params.runCell("codex");
   const drift = classifyRuntimeParityCells({
     pi: pi.cell,
     codex: codex.cell,