diff --git a/native-lib/node/src/result.ts b/native-lib/node/src/result.ts
index 267fe2f..636baa8 100644
--- a/native-lib/node/src/result.ts
+++ b/native-lib/node/src/result.ts
@@ -1,5 +1,74 @@
 import type { ExecutionResult, StreamingResult } from "./types";
 
+/** Normalizes a charset name for comparison: lowercased, non-alphanumerics stripped. */
+function normalizeCharset(charset: string): string {
+  return charset.toLowerCase().replace(/[^a-z0-9]/g, "");
+}
+
+/**
+ * Decodes payload bytes to a string using a DataWeave/IANA charset name.
+ *
+ * The native runtime reports charsets using IANA-style names (e.g. `UTF-16`,
+ * `US-ASCII`, `ISO-8859-1`) that Node's `Buffer.toString` does not accept —
+ * passing them through directly throws `ERR_UNKNOWN_ENCODING`. This maps the
+ * common cases to a valid Node {@link BufferEncoding} and handles UTF-16
+ * byte-order (Node only decodes little-endian): a leading BOM, or a `UTF-16BE`
+ * label, is honored by stripping the BOM and byte-swapping big-endian input.
+ * Unrecognized charsets fall back to UTF-8 so decoding degrades gracefully
+ * instead of throwing.
+ *
+ * @param bytes - The decoded payload bytes.
+ * @param charset - The charset name reported by the runtime, or `null`.
+ * @returns The decoded string.
+ */
+export function decodeBytes(bytes: Buffer, charset: string | null): string {
+  if (!charset) return bytes.toString("utf-8");
+
+  switch (normalizeCharset(charset)) {
+    case "utf16":
+    case "utf16le":
+    case "utf16be":
+    case "ucs2":
+    case "unicode":
+      return decodeUtf16(bytes, normalizeCharset(charset) === "utf16be");
+    case "usascii":
+    case "ascii":
+      return bytes.toString("ascii");
+    case "latin1":
+    case "iso88591":
+    case "cp1252":
+    case "windows1252":
+      return bytes.toString("latin1");
+    case "base64":
+      return bytes.toString("base64");
+    case "hex":
+      return bytes.toString("hex");
+    case "utf8":
+    default:
+      return bytes.toString("utf-8");
+  }
+}
+
+/**
+ * Decodes UTF-16 bytes to a string, honoring a BOM if present (which overrides
+ * `labelIsBE`) and byte-swapping big-endian input, since Node only has a
+ * little-endian UTF-16 decoder.
+ */
+function decodeUtf16(bytes: Buffer, labelIsBE: boolean): string {
+  let big = labelIsBE;
+  let start = 0;
+  if (bytes.length >= 2) {
+    if (bytes[0] === 0xfe && bytes[1] === 0xff) { big = true; start = 2; }       // BE BOM
+    else if (bytes[0] === 0xff && bytes[1] === 0xfe) { big = false; start = 2; } // LE BOM
+  }
+  let body = bytes.subarray(start);
+  if (big) {
+    body = Buffer.from(body); // copy so swap16 doesn't mutate the caller's bytes
+    if (body.length % 2 === 0) body.swap16();
+  }
+  return body.toString("utf16le");
+}
+
 /**
  * Parses the JSON envelope returned by the native `run_script` FFI call into an
  * {@link ExecutionResult}.
@@ -82,7 +151,7 @@ export function makeResult(
       if (!this.success || this.result === null) return null;
       if (this.binary) return this.result;
       const bytes = Buffer.from(this.result, "base64");
-      return bytes.toString((this.charset as BufferEncoding) ?? "utf-8");
+      return decodeBytes(bytes, this.charset);
     },
   };
 }
diff --git a/native-lib/node/tests/integration/edge-cases.test.ts b/native-lib/node/tests/integration/edge-cases.test.ts
new file mode 100644
index 0000000..cfb385e
--- /dev/null
+++ b/native-lib/node/tests/integration/edge-cases.test.ts
@@ -0,0 +1,202 @@
+// Integration edge cases beyond the happy-path suite in dataweave.test.ts:
+// concurrency / FFI lifecycle, and the error & edge-input matrix. These run
+// against the real native library (dwlib). (Closes gaps F3 and F5 in
+// native-lib/node/TESTING_ASSESSMENT.md.)
+import { describe, it, expect, afterAll } from "vitest";
+import { DataWeave, run, runStreaming, runTransform, cleanup } from "../../src/index";
+import type { StreamingResult } from "../../src/types";
+
+afterAll(() => {
+  cleanup();
+});
+
+/** Drains a streaming/transform generator, returning its chunks and terminal metadata. */
+async function drain(gen: AsyncGenerator<Buffer, StreamingResult, undefined>) {
+  const chunks: Buffer[] = [];
+  let r = await gen.next();
+  while (!r.done) {
+    chunks.push(r.value);
+    r = await gen.next();
+  }
+  return { text: Buffer.concat(chunks).toString("utf-8"), chunks, metadata: r.value };
+}
+
+// --- Concurrency / FFI edge cases ----------------------------------------
+
+describe("runTransform with async-iterable input", () => {
+  it("consumes an async generator (pre-buffered read path)", async () => {
+    async function* asyncChunks(): AsyncGenerator<Buffer> {
+      for (const s of ["[1, 2, ", "3, 4", ", 5]"]) yield Buffer.from(s);
+    }
+    const { text, metadata } = await drain(
+      runTransform("output application/json\n---\npayload map ($ * 10)", asyncChunks(), {
+        mimeType: "application/json",
+      })
+    );
+    expect(metadata.success).toBe(true);
+    expect(text).toContain("10");
+    expect(text).toContain("50");
+  });
+
+  it("handles an empty async input", async () => {
+    async function* empty(): AsyncGenerator<Buffer> { /* yields nothing */ }
+    const { metadata } = await drain(
+      runTransform("output application/json\n---\n{ ok: true }", empty(), { mimeType: "application/json" })
+    );
+    // Script ignores payload; empty input must not hang or crash.
+    expect(metadata.success).toBe(true);
+  });
+
+  it("treats a throwing async input as end-of-input for what was produced", async () => {
+    async function* boom(): AsyncGenerator<Buffer> {
+      yield Buffer.from("[1, 2, 3]");
+      throw new Error("source failed");
+    }
+    const { text, metadata } = await drain(
+      runTransform("output application/json\n---\nsizeOf(payload)", boom(), { mimeType: "application/json" })
+    );
+    expect(metadata.success).toBe(true);
+    expect(text.trim()).toBe("3");
+  });
+});
+
+describe("multi-instance lifecycle", () => {
+  it("runs two independent instances and cleans them up independently", () => {
+    const a = new DataWeave();
+    const b = new DataWeave();
+    a.initialize();
+    b.initialize();
+    try {
+      expect(a.run("1 + 1").getString()).toBe("2");
+      expect(b.run("2 + 3").getString()).toBe("5");
+    } finally {
+      a.cleanup();
+      b.cleanup();
+    }
+    // After cleanup, a fresh instance still works (runtime not permanently torn down).
+    const c = new DataWeave();
+    c.initialize();
+    try {
+      expect(c.run("6 * 7").getString()).toBe("42");
+    } finally {
+      c.cleanup();
+    }
+  });
+
+  it("initialize is idempotent and re-initialization after cleanup works", () => {
+    const dw = new DataWeave();
+    dw.initialize();
+    dw.initialize(); // no-op, must not throw
+    expect(dw.run("1").getString()).toBe("1");
+    dw.cleanup();
+    dw.cleanup(); // double cleanup, must not throw
+    dw.initialize(); // re-init
+    try {
+      expect(dw.run("2").getString()).toBe("2");
+    } finally {
+      dw.cleanup();
+    }
+  });
+
+  it("run before initialize throws a DataWeaveError", () => {
+    const dw = new DataWeave();
+    expect(() => dw.run("1 + 1")).toThrow(/not initialized/i);
+  });
+
+  // NOTE: initialize() with a bad library path throwing a DataWeaveError is only
+  // reliably observable as the FIRST initialization in a process — the native
+  // runtime is loaded process-globally (ref-counted, see g_ref_count in
+  // addon.c), so once any good init has run, a later bad-path init is tolerated
+  // rather than re-dlopen'd. Because this shared suite loads dwlib in earlier
+  // tests, that path can't be asserted here without a dedicated isolated process
+  // (a follow-up: run it via a separate vitest pool/child so no prior init has
+  // happened). Verified manually: in a fresh process it throws DataWeaveError.
+});
+
+describe("concurrent execution", () => {
+  it("runs many concurrent scripts on the shared singleton", async () => {
+    const results = await Promise.all(
+      Array.from({ length: 25 }, (_, i) => Promise.resolve().then(() => run(`${i} * 2`)))
+    );
+    results.forEach((r, i) => {
+      expect(r.success).toBe(true);
+      expect(r.getString()).toBe(String(i * 2));
+    });
+  });
+
+  it("interleaves multiple concurrent streaming generators", async () => {
+    const gens = Array.from({ length: 4 }, (_, i) =>
+      runStreaming(`output application/json --- (1 to 200) map ($ + ${i * 1000})`)
+    );
+    const drained = await Promise.all(gens.map(drain));
+    drained.forEach(({ metadata, text }, i) => {
+      expect(metadata.success).toBe(true);
+      expect(text).toContain(String(i * 1000 + 1));
+    });
+  });
+});
+
+// --- Error & edge-input matrix -------------------------------------------
+
+describe("binary output", () => {
+  it("marks binary output and round-trips through getBytes", () => {
+    const r = run("output application/octet-stream\n---\npayload", {
+      payload: { content: Buffer.from([0, 1, 2, 255]), mimeType: "application/octet-stream" },
+    });
+    expect(r.success).toBe(true);
+    expect(r.binary).toBe(true);
+    expect(r.getBytes()!.equals(Buffer.from([0, 1, 2, 255]))).toBe(true);
+  });
+});
+
+describe("output charsets (buffered path)", () => {
+  // Regression guard for the getString() charset bug: the runtime reports IANA
+  // charset names that Node's Buffer.toString rejects; decodeBytes normalizes them.
+  for (const enc of ["UTF-8", "UTF-16", "UTF-16LE", "UTF-16BE", "ISO-8859-1"]) {
+    it(`decodes ${enc} output without throwing`, () => {
+      const r = run(`output text/plain encoding="${enc}"\n---\n"café"`);
+      expect(r.success).toBe(true);
+      expect(() => r.getString()).not.toThrow();
+      expect(r.getString()).toBe("café");
+    });
+  }
+
+  it("substitutes unrepresentable characters for US-ASCII output", () => {
+    const r = run(`output text/plain encoding="US-ASCII"\n---\n"café"`);
+    expect(r.success).toBe(true);
+    // 'é' is not representable in ASCII; the runtime substitutes it on encode.
+    expect(r.getString()).toBe("caf?");
+  });
+});
+
+describe("malformed / unsupported input", () => {
+  it("fails on an unknown input mime type", () => {
+    const r = run("payload", { payload: { content: "x", mimeType: "application/nonsense-xyz" } });
+    expect(r.success).toBe(false);
+    expect(r.error).toBeTruthy();
+  });
+
+  it("fails on an unknown output mime type", () => {
+    const r = run("output application/nonsense-xyz\n---\n{ a: 1 }");
+    expect(r.success).toBe(false);
+    expect(r.error).toBeTruthy();
+  });
+
+  it("fails when referencing a missing input", () => {
+    const r = run("missingInput + 1");
+    expect(r.success).toBe(false);
+    expect(r.error).toBeTruthy();
+  });
+
+  it("fails on a script with a compilation/syntax error", () => {
+    const r = run("output application/json\n---\n{ unclosed:");
+    expect(r.success).toBe(false);
+    expect(r.error).toBeTruthy();
+  });
+
+  it("surfaces malformed input content as a failed result", () => {
+    const r = run("payload.value", { payload: { content: "{ not valid json", mimeType: "application/json" } });
+    expect(r.success).toBe(false);
+    expect(r.error).toBeTruthy();
+  });
+});
\ No newline at end of file
diff --git a/native-lib/node/tests/unit/result.test.ts b/native-lib/node/tests/unit/result.test.ts
index 4d9d6d6..3ceb423 100644
--- a/native-lib/node/tests/unit/result.test.ts
+++ b/native-lib/node/tests/unit/result.test.ts
@@ -1,5 +1,5 @@
 import { describe, it, expect } from "vitest";
-import { parseNativeResponse, makeResult, parseStreamingResult } from "../../src/result";
+import { parseNativeResponse, makeResult, parseStreamingResult, decodeBytes } from "../../src/result";
 
 const b64 = (s: string, enc: BufferEncoding = "utf-8") =>
   Buffer.from(s, enc).toString("base64");
@@ -98,6 +98,61 @@ describe("makeResult", () => {
   });
 });
 
+describe("decodeBytes", () => {
+  it("defaults to UTF-8 when charset is null", () => {
+    expect(decodeBytes(Buffer.from("café", "utf-8"), null)).toBe("café");
+  });
+
+  it("decodes UTF-8 by name", () => {
+    expect(decodeBytes(Buffer.from("café", "utf-8"), "UTF-8")).toBe("café");
+  });
+
+  it("decodes little-endian UTF-16", () => {
+    const le = Buffer.from("café", "utf16le");
+    expect(decodeBytes(le, "UTF-16LE")).toBe("café");
+  });
+
+  it("decodes big-endian UTF-16 by label (byte-swapping)", () => {
+    const be = Buffer.from("café", "utf16le");
+    be.swap16();
+    expect(decodeBytes(be, "UTF-16BE")).toBe("café");
+  });
+
+  it("honors a big-endian BOM regardless of label", () => {
+    const body = Buffer.from("café", "utf16le");
+    body.swap16();
+    const beWithBom = Buffer.concat([Buffer.from([0xfe, 0xff]), body]);
+    expect(decodeBytes(beWithBom, "UTF-16")).toBe("café");
+  });
+
+  it("honors a little-endian BOM and strips it", () => {
+    const leWithBom = Buffer.concat([Buffer.from([0xff, 0xfe]), Buffer.from("café", "utf16le")]);
+    expect(decodeBytes(leWithBom, "UTF-16")).toBe("café");
+  });
+
+  it("decodes ISO-8859-1 / latin1", () => {
+    const latin = Buffer.from("café", "latin1");
+    expect(decodeBytes(latin, "ISO-8859-1")).toBe("café");
+  });
+
+  it("decodes US-ASCII", () => {
+    expect(decodeBytes(Buffer.from("hello", "ascii"), "US-ASCII")).toBe("hello");
+  });
+
+  it("falls back to UTF-8 for an unrecognized charset instead of throwing", () => {
+    expect(() => decodeBytes(Buffer.from("hi", "utf-8"), "x-made-up-charset")).not.toThrow();
+    expect(decodeBytes(Buffer.from("hi", "utf-8"), "x-made-up-charset")).toBe("hi");
+  });
+
+  it("does not mutate the caller's buffer when byte-swapping", () => {
+    const be = Buffer.from("AB", "utf16le");
+    be.swap16();
+    const snapshot = Buffer.from(be);
+    decodeBytes(be, "UTF-16BE");
+    expect(be.equals(snapshot)).toBe(true);
+  });
+});
+
 describe("parseStreamingResult", () => {
   it("treats an empty string as a failure", () => {
     const r = parseStreamingResult("");