diff --git a/frontend/src/pages/Findings.tsx b/frontend/src/pages/Findings.tsx
index 61e6d44e4..39050aa03 100644
--- a/frontend/src/pages/Findings.tsx
+++ b/frontend/src/pages/Findings.tsx
@@ -602,6 +602,25 @@ export default function Findings() {
   }
 
   async function loadMore() {
+    if (loadingMore) return
+    setLoadingMore(true)
+    const nextPage = page + 1
+    try {
+      const data = await getFindings(nextPage, perPage)
+      const rawFindings = data.findings || []
+      const moreFindings = rawFindings.filter(
+        (finding) => typeof finding.id === 'string',
+      ) as Finding[]
+      if (rawFindings.length > 0) {
+        setFindings((prev) => [...prev, ...moreFindings])
+        setPage(nextPage)
+      }
+      // Fix #1862: keep totalItems in sync with each /findings response so
+      // the "Load More" guard (findings.length < totalItems) stays accurate
+      // even when filters change the server-side total between pages.
+      setTotalItems(data.total ?? moreFindings.length)
+    } finally {
+      setLoadingMore(false)
   if (loadingMore) return
   setLoadingMore(true)
   const nextPage = page + 1
diff --git a/frontend/testing/unit/pages/Findings.test.tsx b/frontend/testing/unit/pages/Findings.test.tsx
index 596fce909..7b5b5563a 100644
--- a/frontend/testing/unit/pages/Findings.test.tsx
+++ b/frontend/testing/unit/pages/Findings.test.tsx
@@ -537,6 +537,100 @@ describe('Findings — virtualizer scrolling', () => {
 })
 
 it('scrolls to the correct fresh index after sort order changes then selection changes', async () => {
+  const findings = [
+    makeFinding({ id: 'f1', title: 'Finding Alpha', severity: 'critical', discovered_at: '2024-01-01T00:00:00Z' }),
+    makeFinding({ id: 'f2', title: 'Finding Beta', severity: 'high', discovered_at: '2024-01-03T00:00:00Z' }),
+    makeFinding({ id: 'f3', title: 'Finding Gamma', severity: 'medium', discovered_at: '2024-01-02T00:00:00Z' }),
+  ]
+  vi.mocked(getFindings).mockResolvedValue({ findings })
+
+  render(<Findings />)
+  await waitFor(() => expect(screen.queryByText('Synchronizing findings feed...')).not.toBeInTheDocument())
+
+  // Switch to "newest" sort — new order is Beta(0), Gamma(1), Alpha(2)
+  const selects = screen.getAllByRole('combobox')
+  const sortSelect = selects.find((s) =>
+    Array.from(s.querySelectorAll('option')).some((o) => /Newest First/i.test(o.textContent || '')),
+  )
+  await userEvent.selectOptions(sortSelect!, 'newest')
+
+  mockScrollToIndex.mockClear()
+
+  // Now select Gamma — should scroll to its *post-sort* index (1), not a stale pre-sort index
+  const gammaOption = await screen.findByRole('option', { name: /Finding Gamma/i })
+  await userEvent.click(gammaOption)
+
+  expect(mockScrollToIndex).toHaveBeenCalledWith(1, { align: 'auto', behavior: 'smooth' })
+})
+
+describe('Findings — load more totalItems sync (#1862)', () => {
+  beforeEach(() => {
+    vi.clearAllMocks()
+    localStorage.clear()
+  })
+
+  it('updates totalItems after each loadMore fetch so the button guard stays accurate', async () => {
+    // Initial load: 2 findings, server reports 10 total
+    const page1 = [
+      makeFinding({ id: 'p1-f1', title: 'Page 1 Finding A' }),
+      makeFinding({ id: 'p1-f2', title: 'Page 1 Finding B' }),
+    ]
+    // loadMore call: 2 more findings, server now reports total=4 (filter narrowed)
+    const page2 = [
+      makeFinding({ id: 'p2-f1', title: 'Page 2 Finding A' }),
+      makeFinding({ id: 'p2-f2', title: 'Page 2 Finding B' }),
+    ]
+
+    vi.mocked(getFindings)
+      .mockResolvedValueOnce({ findings: page1, total: 10 })
+      .mockResolvedValueOnce({ findings: page2, total: 4 })
+
+    render(<Findings />)
+    await waitFor(() =>
+      expect(screen.queryByText('Synchronizing findings feed...')).not.toBeInTheDocument(),
+    )
+
+    // After initial load: 2 findings loaded, server total=10, button shows "Load More (2/10)"
+    const loadMoreBtn = screen.getByRole('button', { name: /Load More/i })
+    expect(loadMoreBtn).toHaveTextContent('Load More (2/10)')
+
+    // Click Load More — triggers second fetch (total updates to 4)
+    await userEvent.click(loadMoreBtn)
+
+    // After loadMore: 4 findings loaded, totalItems updated to 4 → button hidden (4 >= 4)
+    await waitFor(() =>
+      expect(screen.queryByRole('button', { name: /Load More/i })).not.toBeInTheDocument(),
+    )
+  })
+
+  it('shows Load More button when loadMore response total exceeds current findings count', async () => {
+    const page1 = [
+      makeFinding({ id: 'p1-f1', title: 'Page 1 Finding' }),
+    ]
+    const page2 = [
+      makeFinding({ id: 'p2-f1', title: 'Page 2 Finding' }),
+    ]
+
+    vi.mocked(getFindings)
+      .mockResolvedValueOnce({ findings: page1, total: 5 })
+      .mockResolvedValueOnce({ findings: page2, total: 5 })
+
+    render(<Findings />)
+    await waitFor(() =>
+      expect(screen.queryByText('Synchronizing findings feed...')).not.toBeInTheDocument(),
+    )
+
+    // Initial state: 1/5 loaded, button visible
+    expect(screen.getByRole('button', { name: /Load More \(1\/5\)/i })).toBeInTheDocument()
+
+    await userEvent.click(screen.getByRole('button', { name: /Load More/i }))
+
+    // After loadMore: 2/5, totalItems stays 5, button still visible
+    await waitFor(() =>
+      expect(screen.getByRole('button', { name: /Load More \(2\/5\)/i })).toBeInTheDocument(),
+    )
+  })
+})
     const findings = [
       makeFinding({ id: 'f1', title: 'Finding Alpha', severity: 'critical', discovered_at: '2024-01-01T00:00:00Z' }),
       makeFinding({ id: 'f2', title: 'Finding Beta', severity: 'high', discovered_at: '2024-01-03T00:00:00Z' }),
diff --git a/testing/backend/unit/test_semgrep_scanner_plugin.py b/testing/backend/unit/test_semgrep_scanner_plugin.py
index 032f2c938..8814a438e 100644
--- a/testing/backend/unit/test_semgrep_scanner_plugin.py
+++ b/testing/backend/unit/test_semgrep_scanner_plugin.py
@@ -114,3 +114,80 @@ def test_semgrep_parser_severity_mapping():
 
         parsed = parser.parse(json_data)
         assert parsed["findings"][0]["severity"] == expected_secuscan_sev
+
+
+class TestSemgrepParserMalformedJsonFallback:
+    """
+    Verify the Semgrep parser's silent fallback behaviour when JSON is malformed.
+
+    The parser wraps all parsing in ``except Exception: pass``, so every
+    malformed-input variant must deterministically return
+    ``{"count": 0, "findings": []}``.
+    """
+
+    def test_truncated_json_returns_empty_findings(self):
+        """Truncated JSON (open object never closed) must return count=0, findings=[].
+
+        Simulates a scanner process that was killed mid-write, leaving an
+        incomplete JSON payload in stdout.
+        """
+        parser = _load_semgrep_parser()
+        truncated = '{"results": [{'
+
+        parsed = parser.parse(truncated)
+
+        assert parsed["count"] == 0
+        assert parsed["findings"] == []
+
+    def test_mixed_stdout_with_json_fragment_returns_deterministic_empty_result(self):
+        """Mixed stdout containing log lines + a JSON fragment must return count=0.
+
+        Real Semgrep invocations may print warning/info lines to stdout before
+        the JSON block.  If the full stdout is fed to the parser the result
+        must still be a deterministic empty-findings dict, not a crash.
+        """
+        parser = _load_semgrep_parser()
+        mixed_stdout = (
+            "Running semgrep...\n"
+            "Loading rules from registry...\n"
+            '{"results": [{"check_id": "rule-x"'  # fragment — never closed
+        )
+
+        parsed = parser.parse(mixed_stdout)
+
+        assert parsed["count"] == 0
+        assert parsed["findings"] == []
+        # Call twice to confirm determinism
+        parsed_again = parser.parse(mixed_stdout)
+        assert parsed_again["count"] == 0
+        assert parsed_again["findings"] == []
+
+    def test_valid_json_missing_top_level_results_key_returns_empty(self):
+        """Valid JSON that lacks the top-level ``results`` key must return count=0.
+
+        The parser calls ``data.get("results", [])``, so missing the key
+        should yield an empty findings list rather than raise.
+        """
+        parser = _load_semgrep_parser()
+        no_results_key = json.dumps({"version": "1.0", "errors": []})
+
+        parsed = parser.parse(no_results_key)
+
+        assert parsed["count"] == 0
+        assert parsed["findings"] == []
+
+    def test_valid_json_with_null_in_critical_fields_returns_empty_without_crash(self):
+        """Null values in critical fields must not raise and must return count=0.
+
+        Some Semgrep builds (or mocked environments) can emit ``null`` for
+        ``results`` itself.  The parser must absorb this gracefully because
+        ``null`` is valid JSON but iteration over it raises ``TypeError``,
+        which the broad ``except Exception`` clause catches.
+        """
+        parser = _load_semgrep_parser()
+        null_results = json.dumps({"results": None})
+
+        parsed = parser.parse(null_results)
+
+        assert parsed["count"] == 0
+        assert parsed["findings"] == []