diff --git a/packages/email-ingestion-gateway/package.json b/packages/email-ingestion-gateway/package.json
index 66b71b67b..26c23e5ee 100644
--- a/packages/email-ingestion-gateway/package.json
+++ b/packages/email-ingestion-gateway/package.json
@@ -54,7 +54,7 @@
     "graphql-request": "7.4.0",
     "inline-css": "4.0.3",
     "playwright": "1.62.1",
-    "postal-mime": "2.7.6",
+    "postal-mime": "3.0.0",
     "zod": "4.4.3"
   },
   "devDependencies": {
diff --git a/packages/email-ingestion-gateway/src/__tests__/mime-extractor.test.ts b/packages/email-ingestion-gateway/src/__tests__/mime-extractor.test.ts
index ab4744942..3bc38b509 100644
--- a/packages/email-ingestion-gateway/src/__tests__/mime-extractor.test.ts
+++ b/packages/email-ingestion-gateway/src/__tests__/mime-extractor.test.ts
@@ -444,6 +444,81 @@ describe('extractFromMime — senderEvidence', () => {
     expect(result.senderEvidence.originalSender).toBe('orig-sender@vendor.com');
   });
 
+  // postal-mime v3 unfolds long headers by dropping the CRLF only, leaving the fold's
+  // space/tab inside the value (v2 collapsed every whitespace run itself). Everything
+  // downstream — the server's `' via '` split, its display-name comparison against the
+  // tenant's own business names — assumes single spaces, so the extractor normalizes.
+  it('collapses the whitespace RFC 5322 folding leaves in header values', async () => {
+    const mime = Buffer.from(
+      [
+        'From: sender@example.com',
+        'To: invoices@example.com',
+        'Subject: Test',
+        'X-Original-From: "Vendor Ltd" via',
+        '\tAccount Payables <billing@vendor.com>',
+        'Content-Type: text/plain',
+        '',
+        'body',
+      ].join('\r\n'),
+      'utf8',
+    );
+    const result = (await extractFromMime(mime)) as { success: true; senderEvidence: SenderEvidence };
+    expect(result.senderEvidence.originalFrom).toBe(
+      '"Vendor Ltd" via Account Payables <billing@vendor.com>',
+    );
+  });
+
+  it('collapses folding whitespace inside the From display name', async () => {
+    const mime = Buffer.from(
+      [
+        'From: "Vendor Ltd" via',
+        '\tAccount Payables <group@tenant.example>',
+        'To: invoices@example.com',
+        'Subject: Test',
+        'Content-Type: text/plain',
+        '',
+        'body',
+      ].join('\r\n'),
+      'utf8',
+    );
+    const result = (await extractFromMime(mime)) as { success: true; senderEvidence: SenderEvidence };
+    expect(result.senderEvidence.fromDisplayName).toBe('Vendor Ltd via Account Payables');
+    expect(result.senderEvidence.from).toBe('Vendor Ltd via Account Payables <group@tenant.example>');
+  });
+
+  it('takes the first Reply-To address as the primary one', async () => {
+    const mime = makeMultipartMime({
+      replyTo: 'primary@vendor.com, secondary@vendor.com',
+      attachments: [{ filename: 'a.pdf', mimeType: 'application/pdf', content: fakePdf() }],
+    });
+    const result = (await extractFromMime(mime)) as { success: true; senderEvidence: SenderEvidence };
+    expect(result.senderEvidence.replyTo).toBe('primary@vendor.com');
+  });
+
+  it('resolves a duplicated single-value header to its first occurrence', async () => {
+    const mime = Buffer.from(
+      [
+        'From: sender@example.com',
+        'To: invoices@example.com',
+        'Subject: First',
+        'Subject: Second',
+        'X-Original-Sender: first@vendor.com',
+        'X-Original-Sender: second@vendor.com',
+        'Content-Type: text/plain',
+        '',
+        'body',
+      ].join('\r\n'),
+      'utf8',
+    );
+    const result = (await extractFromMime(mime)) as {
+      success: true;
+      subject: string;
+      senderEvidence: SenderEvidence;
+    };
+    expect(result.subject).toBe('First');
+    expect(result.senderEvidence.originalSender).toBe('first@vendor.com');
+  });
+
   it('returns undefined for absent optional headers', async () => {
     const mime = makeMultipartMime({
       attachments: [{ filename: 'a.pdf', mimeType: 'application/pdf', content: fakePdf() }],
diff --git a/packages/email-ingestion-gateway/src/mime-extractor.ts b/packages/email-ingestion-gateway/src/mime-extractor.ts
index ca871fac8..8f0af261c 100644
--- a/packages/email-ingestion-gateway/src/mime-extractor.ts
+++ b/packages/email-ingestion-gateway/src/mime-extractor.ts
@@ -131,7 +131,9 @@ export async function extractFromMime(rawMime: Buffer): Promise<ExtractionResult
     body,
     senderEvidence: {
       from: formatAddress(email.from),
-      fromDisplayName: email.from?.name?.trim() || undefined,
+      fromDisplayName: unfold(email.from?.name) || undefined,
+      // postal-mime v3 returns address lists in document order (v2 reversed them),
+      // so [0] is the header's first — i.e. primary — Reply-To address.
       replyTo: formatAddress(email.replyTo?.[0]),
       // Kept as separate fields: X-Original-From is the original *author* (often a
       // display name only), X-Original-Sender the relaying platform. Reading them as
@@ -194,10 +196,25 @@ function toBuffer(content: ArrayBuffer | Uint8Array | string): Buffer {
   return Buffer.from(new Uint8Array(content));
 }
 
+/**
+ * Collapse every whitespace run to a single space and trim.
+ *
+ * Header folding (RFC 5322) inserts `CRLF` + a space **or tab** at arbitrary points in
+ * a long header, and postal-mime v3 unfolds by dropping only the `CRLF` — so the fold's
+ * tab (or run of spaces) survives inside the value and inside display names parsed out
+ * of it. v2 collapsed those runs itself; everything reading this evidence assumes single
+ * spaces (`splitViaDisplayName` looks for a literal `' via '`, the server compares display
+ * names against the tenant's own business names), so normalize here rather than teaching
+ * every consumer about folding.
+ */
+function unfold(value: string | null | undefined): string {
+  return value ? value.replaceAll(/\s+/g, ' ').trim() : '';
+}
+
 /** Render a parsed address as `Name <addr>` (or just the address / display name). */
 function formatAddress(addr: Address | undefined): string | undefined {
   if (!addr) return undefined;
-  const name = addr.name?.trim();
+  const name = unfold(addr.name);
   const address = addr.address?.trim();
   if (name && address) return `${name} <${address}>`;
   return address || name || undefined;
@@ -213,12 +230,15 @@ function headerValue(
 ): string | undefined {
   for (const name of names) {
     const lower = name.toLowerCase();
+    // First occurrence wins, matching postal-mime v3's own duplicate-header resolution
+    // for single-value headers such as `subject` / `from`.
     const found = headers.find(header => header.key.toLowerCase() === lower);
-    if (found?.value) {
+    const value = unfold(found?.value);
+    if (value) {
       try {
-        return decodeWords(found.value);
+        return decodeWords(value);
       } catch {
-        return found.value;
+        return value;
       }
     }
   }
diff --git a/yarn.lock b/yarn.lock
index 8af2aa8a7..d79a1e966 100644
--- a/yarn.lock
+++ b/yarn.lock
@@ -149,7 +149,7 @@ __metadata:
     graphql-request: "npm:7.4.0"
     inline-css: "npm:4.0.3"
     playwright: "npm:1.62.1"
-    postal-mime: "npm:2.7.6"
+    postal-mime: "npm:3.0.0"
     tsx: "npm:4.23.12"
     typescript: "npm:6.0.3"
     wrangler: "npm:4.122.0"
@@ -23248,10 +23248,10 @@ __metadata:
   languageName: node
   linkType: hard
 
-"postal-mime@npm:2.7.6":
-  version: 2.7.6
-  resolution: "postal-mime@npm:2.7.6"
-  checksum: 10c0/7932b9b5c82e9d3f25607d9f07eba92414b42a304592f8cd03f9df058311af8d5df091ea8ef05ce5f20f052cbd858bdda8df0903d547a54953814ca57a68df03
+"postal-mime@npm:3.0.0":
+  version: 3.0.0
+  resolution: "postal-mime@npm:3.0.0"
+  checksum: 10c0/e7385b3ceccba9e14a0c81d46961ab5198d8bf12ce7a43925349deb7d38d647d319555379c70d1bfc437bb0e765cae062cac3105a7092d5b676223749d0d9010
   languageName: node
   linkType: hard