diff --git a/packages/frontend/tests/dataProcessing.stress.test.ts b/packages/frontend/tests/dataProcessing.stress.test.ts
new file mode 100644
index 0000000..b28478d
--- /dev/null
+++ b/packages/frontend/tests/dataProcessing.stress.test.ts
@@ -0,0 +1,220 @@
+// Frontend-layer stress guards: exercise @jspsych/metadata's generate() and the browser
+// validatePsychDS wrapper through the frontend's own dependency resolution. The scale / type-
+// inference / many-levels blocks deliberately overlap packages/metadata's *.stress suites — here
+// they guard the generate() the *frontend bundle* actually loads (a different resolution path),
+// alongside the genuinely frontend-only coverage: multi-file CSV+JSON accumulation and the
+// validator error/warning partitioning (Smoke-test-2 regression).
+import JsPsychMetadata from "@jspsych/metadata";
+import { validateWeb } from "psychds-validator/web/psychds-validator.js";
+import { validatePsychDS } from "../src/validation/validatePsychDS";
+import { validatorOutput } from "./helpers";
+
+const mockValidateWeb = validateWeb as jest.Mock;
+
+interface VariableMeta {
+  minValue?: number;
+  maxValue?: number;
+  levels?: unknown[];
+  value?: string;
+}
+
+// Mock PluginCache npm fetches — ok:true prevents the "source not found" warn path.
+const mockFetch = jest.fn().mockResolvedValue({ ok: true, text: () => Promise.resolve("") });
+beforeEach(() => {
+  (global as typeof globalThis & { fetch: jest.Mock }).fetch = mockFetch;
+  mockFetch.mockClear();
+  mockValidateWeb.mockReset();
+});
+afterEach(() => jest.restoreAllMocks());
+
+// Build a CSV string from an array of row objects.
+function makeCsv(rows: Record<string, unknown>[]): string {
+  if (rows.length === 0) return "";
+  const headers = Object.keys(rows[0]);
+  const escape = (v: unknown) => {
+    if (v === null || v === undefined) return "null";
+    const s = String(v);
+    return s.includes(",") || s.includes('"') ? `"${s.replace(/"/g, '""')}"` : s;
+  };
+  return [headers.join(","), ...rows.map((r) => headers.map((h) => escape(r[h])).join(","))].join("\n");
+}
+
+function baseRow(overrides: Record<string, unknown> = {}) {
+  return { trial_type: "html-keyboard-response", trial_index: 0, time_elapsed: 500, ...overrides };
+}
+
+// ─── Scale: large CSV ─────────────────────────────────────────────────────────
+
+describe("Scale: large CSV (1,000 rows)", () => {
+  test("detects all 7 variables", async () => {
+    const rows = Array.from({ length: 1000 }, (_, i) =>
+      baseRow({ trial_index: i, time_elapsed: i * 100, stimulus: `s${i}`, response: i % 2 ? "f" : "j", rt: 400 + i, correct: i % 3 !== 0 }),
+    );
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(rows), {}, "csv");
+    const names = meta.getVariableNames();
+    for (const col of ["trial_type", "trial_index", "time_elapsed", "stimulus", "response", "rt", "correct"]) {
+      expect(names).toContain(col);
+    }
+  });
+
+  test("computes correct numeric range across 1,000 rows", async () => {
+    const rows = Array.from({ length: 1000 }, (_, i) => baseRow({ trial_index: i, rt: i + 1 }));
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(rows), {}, "csv");
+    const rt = meta.getVariable("rt") as VariableMeta;
+    expect(rt.minValue).toBe(1);
+    expect(rt.maxValue).toBe(1000);
+  });
+});
+
+// ─── Scale: wide CSV ─────────────────────────────────────────────────────────
+
+describe("Scale: wide CSV (31 columns)", () => {
+  test("detects all 31 columns", async () => {
+    const cols = Array.from({ length: 30 }, (_, i) => `col_${i}`);
+    const rows = Array.from({ length: 5 }, (_, i) =>
+      Object.fromEntries([["trial_index", i], ...cols.map((c) => [c, i])]),
+    );
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(rows), {}, "csv");
+    const names = meta.getVariableNames();
+    expect(names).toContain("trial_index");
+    for (const col of cols) expect(names).toContain(col);
+  });
+});
+
+// ─── Multi-file: CSV + JSON accumulation ─────────────────────────────────────
+// The frontend calls generate() once per file. This block guards that shared
+// variables don't get duplicated and JSON-only variables are captured.
+
+describe("Multi-file: CSV + JSON accumulation", () => {
+  test("shared variables appear exactly once after processing both files", async () => {
+    const csvRows = [baseRow({ stimulus: "+", response: "f", rt: 800 })];
+    const jsonRows = [{ ...baseRow({ trial_index: 1 }), stimulus: "cat", response: "j", rt: 900 }];
+
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(csvRows), {}, "csv");
+    await meta.generate(JSON.stringify(jsonRows), {}, "json");
+
+    const names = meta.getVariableNames();
+    expect(names.filter((n) => n === "trial_type").length).toBe(1);
+    expect(names.filter((n) => n === "rt").length).toBe(1);
+  });
+
+  test("JSON-only variables are added on top of CSV variables", async () => {
+    const csvRows = [baseRow({ response: "f" })];
+    const jsonRows = [{ ...baseRow(), subject: "s01", response: "j" }];
+
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(csvRows), {}, "csv");
+    await meta.generate(JSON.stringify(jsonRows), {}, "json");
+
+    expect(meta.getVariableNames()).toContain("subject");
+  });
+
+  test("numeric range spans values from both files", async () => {
+    const csvRows = [baseRow({ rt: 200 }), baseRow({ trial_index: 1, rt: 400 })];
+    const jsonRows = [{ ...baseRow({ trial_index: 2 }), rt: 100 }, { ...baseRow({ trial_index: 3 }), rt: 800 }];
+
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(csvRows), {}, "csv");
+    await meta.generate(JSON.stringify(jsonRows), {}, "json");
+
+    const rt = meta.getVariable("rt") as VariableMeta;
+    expect(rt.minValue).toBe(100);
+    expect(rt.maxValue).toBe(800);
+  });
+});
+
+// ─── Type inference ──────────────────────────────────────────────────────────
+
+describe("Type inference", () => {
+  test("native boolean values in JSON are typed as boolean", async () => {
+    const data = JSON.stringify([
+      { ...baseRow(), correct: true },
+      { ...baseRow({ trial_index: 1 }), correct: false },
+    ]);
+    const meta = new JsPsychMetadata();
+    await meta.generate(data, {}, "json");
+    expect((meta.getVariable("correct") as VariableMeta).value).toBe("boolean");
+  });
+
+  test("null values in a numeric CSV column do not affect range detection", async () => {
+    const rows = [
+      baseRow({ rt: "null" }),
+      baseRow({ trial_index: 1, rt: 500 }),
+      baseRow({ trial_index: 2, rt: 800 }),
+    ];
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(rows), {}, "csv");
+    const rt = meta.getVariable("rt") as VariableMeta;
+    expect(rt.minValue).toBe(500);
+    expect(rt.maxValue).toBe(800);
+  });
+});
+
+// ─── Many levels: "Show all N" guard ─────────────────────────────────────────
+// The Variables step truncates levels at 5 and shows "Show all N". This guard
+// ensures that all levels are recorded so the UI's count and full list are correct.
+
+describe("Many levels", () => {
+  test("all 50 unique values are stored for a high-cardinality string column", async () => {
+    const rows = Array.from({ length: 50 }, (_, i) => baseRow({ trial_index: i, stimulus: `stimulus-${i}` }));
+    const meta = new JsPsychMetadata();
+    await meta.generate(makeCsv(rows), {}, "csv");
+    const stim = meta.getVariable("stimulus") as VariableMeta;
+    expect(stim.levels).toHaveLength(50);
+  });
+});
+
+// ─── Smoke-test-2 regression: validator layer ─────────────────────────────────
+// Regression guard for the mixed CSV + JSON case. JSON-only variables (subject,
+// response.Q0, response.Q1, element_index, response.value) appear in variableMeasured
+// but not in any CSV column, so the validator must report VARIABLE_MISSING_FROM_CSV_COLUMNS.
+// PR #103 (fix/frontend-missing-datafile) will resolve this; guard remains until merged.
+
+describe("Smoke-test-2 regression: VARIABLE_MISSING_FROM_CSV_COLUMNS", () => {
+  test("JSON-only variables surface as VARIABLE_MISSING_FROM_CSV_COLUMNS", async () => {
+    mockValidateWeb.mockResolvedValue(
+      validatorOutput([
+        {
+          key: "VARIABLE_MISSING_FROM_CSV_COLUMNS",
+          reason: "variable name in variableMeasured not found in any CSV column",
+          severity: "error",
+          evidence: ["subject,element_index,response.value,response.Q0,response.Q1"],
+        },
+      ]),
+    );
+
+    const dataFiles = new Map([
+      ["experiment/subject-01_data.csv", "trial_type,trial_index,time_elapsed,stimulus,response,rt,correct\nhtml-keyboard-response,0,812,+,null,null,null"],
+    ]);
+
+    const result = await validatePsychDS("{}", dataFiles);
+
+    expect(result.valid).toBe(false);
+    expect(result.errors[0].key).toBe("VARIABLE_MISSING_FROM_CSV_COLUMNS");
+    expect(result.errors[0].evidence[0]).toContain("subject");
+    expect(result.errors[0].evidence[0]).toContain("response.Q0");
+  });
+
+  test("INVALID_SCHEMAORG_PROPERTY on .levels surfaces as a warning, not an error", async () => {
+    mockValidateWeb.mockResolvedValue(
+      validatorOutput([
+        {
+          key: "INVALID_SCHEMAORG_PROPERTY",
+          reason: "levels is not a valid schema.org property",
+          severity: "warning",
+          evidence: [".variableMeasured.levels"],
+        },
+      ]),
+    );
+
+    const result = await validatePsychDS("{}");
+
+    expect(result.valid).toBe(true);
+    expect(result.errors).toHaveLength(0);
+    expect(result.warnings[0].key).toBe("INVALID_SCHEMAORG_PROPERTY");
+  });
+});
diff --git a/packages/frontend/tests/helpers.ts b/packages/frontend/tests/helpers.ts
new file mode 100644
index 0000000..8b19b65
--- /dev/null
+++ b/packages/frontend/tests/helpers.ts
@@ -0,0 +1,25 @@
+/** Builds a validator output object in the shape validatePsychDS consumes. */
+export function validatorOutput(
+  issues: {
+    key: string;
+    reason: string;
+    severity: "error" | "warning";
+    evidence?: (string | undefined)[];
+  }[],
+) {
+  return {
+    issues: new Map(
+      issues.map((issue) => [
+        issue.key,
+        {
+          key: issue.key,
+          reason: issue.reason,
+          severity: issue.severity,
+          files: new Map(
+            (issue.evidence ?? []).map((evidence, i) => [`file${i}`, { evidence }]),
+          ),
+        },
+      ]),
+    ),
+  };
+}
diff --git a/packages/frontend/tests/validatePsychDS.test.ts b/packages/frontend/tests/validatePsychDS.test.ts
index 7ee656d..f2f7e7c 100644
--- a/packages/frontend/tests/validatePsychDS.test.ts
+++ b/packages/frontend/tests/validatePsychDS.test.ts
@@ -4,6 +4,7 @@ import {
   validatePsychDS,
   ValidationUnavailableError,
 } from "../src/validation/validatePsychDS";
+import { validatorOutput } from "./helpers";
 
 const mockValidateWeb = validateWeb as jest.Mock;
 
@@ -17,31 +18,6 @@ function blobText(blob: Blob): Promise<string> {
   });
 }
 
-/** Builds a validator output object in the shape validatePsychDS consumes. */
-function validatorOutput(
-  issues: {
-    key: string;
-    reason: string;
-    severity: "error" | "warning";
-    evidence?: (string | undefined)[];
-  }[],
-) {
-  return {
-    issues: new Map(
-      issues.map((issue) => [
-        issue.key,
-        {
-          key: issue.key,
-          reason: issue.reason,
-          severity: issue.severity,
-          files: new Map(
-            (issue.evidence ?? []).map((evidence, i) => [`file${i}`, { evidence }]),
-          ),
-        },
-      ]),
-    ),
-  };
-}
 
 beforeEach(() => {
   mockValidateWeb.mockReset();