diff --git a/.gitignore b/.gitignore
index 09109dfa2..e347f0b1d 100644
--- a/.gitignore
+++ b/.gitignore
@@ -30,3 +30,6 @@ Thumbs.db
 
 # Pre-commit cache
 .cache
+
+# Downloaded binaries (e.g. pypandoc)
+*.deb
diff --git a/scripts/assets/idf-fields.css b/scripts/assets/idf-fields.css
index fe7fbfcc8..1dacfd5e9 100644
--- a/scripts/assets/idf-fields.css
+++ b/scripts/assets/idf-fields.css
@@ -88,3 +88,41 @@
   border-radius: 3px;
   margin-right: 0.2em;
 }
+
+/* ── Object separator ── */
+hr.idf-object-separator {
+  border: none;
+  border-top: 2px solid var(--md-default-fg-color--lightest);
+  margin: 2.5em 0 1em;
+}
+
+/* ── Sticky object headings ──
+ * Any h2/h3 immediately after the separator sticks to the top of the
+ * viewport while the user scrolls through that object's fields.
+ */
+hr.idf-object-separator + h2,
+hr.idf-object-separator + h3 {
+  position: sticky;
+  top: 0;
+  z-index: 2;
+  background: var(--md-default-bg-color);
+  padding-top: 0.4em;
+  padding-bottom: 0.3em;
+  margin-top: 0;
+  /* subtle bottom edge so the heading doesn't blend into content */
+  box-shadow: 0 1px 0 var(--md-default-fg-color--lightest);
+}
+
+/* When header is hidden on scroll (e.g. Material "header.autohide"),
+ * adjust sticky offset to account for the header height. */
+[data-md-header="shadow"] hr.idf-object-separator + h2,
+[data-md-header="shadow"] hr.idf-object-separator + h3 {
+  top: 0;
+}
+
+/* When the Material header is visible, offset below it.
+ * Material's header is ~3.6rem (default). */
+.md-header--shadow ~ .md-container hr.idf-object-separator + h2,
+.md-header--shadow ~ .md-container hr.idf-object-separator + h3 {
+  top: 0;
+}
diff --git a/scripts/latex_preprocessor.py b/scripts/latex_preprocessor.py
index 49428468a..67e8a0014 100644
--- a/scripts/latex_preprocessor.py
+++ b/scripts/latex_preprocessor.py
@@ -146,34 +146,42 @@ def _find_brace_content(text: str, start: int) -> tuple[str, int] | None:
 }
 
 
+_BRACKET_MACRO_RE = re.compile(r"\\(?:PB|RB|CB)\{")
+
+
 def _expand_all_bracket_macros(text: str) -> str:
     r"""Expand all ``\PB``, ``\RB``, ``\CB`` macros in *text*, inside-out.
 
     When an outer macro wraps inner macros (e.g. ``\PB{a \PB{b}}``) the
     inner content is recursively expanded first, so the final result
     contains no bracket macros regardless of nesting depth.
+
+    Uses regex to jump to the next macro occurrence instead of scanning
+    every character, which is critical for large files.
     """
     result: list[str] = []
-    i = 0
-    while i < len(text):
-        matched_macro = None
-        for macro in _BRACKET_MACROS:
-            if text[i:].startswith(macro + "{"):
-                matched_macro = macro
-                break
-        if matched_macro:
-            brace_start = i + len(matched_macro)
-            found = _find_brace_content(text, brace_start)
-            if found:
-                content, end = found
-                # Recursively expand any bracket macros inside the content
-                content = _expand_all_bracket_macros(content)
-                left, right = _BRACKET_MACROS[matched_macro]
-                result.append(f"{left} {content} {right}")
-                i = end
-                continue
-        result.append(text[i])
-        i += 1
+    pos = 0
+    while True:
+        m = _BRACKET_MACRO_RE.search(text, pos)
+        if m is None:
+            result.append(text[pos:])
+            break
+        # Append everything before this macro
+        result.append(text[pos : m.start()])
+        macro = text[m.start() : m.end() - 1]  # e.g. "\\PB"
+        brace_start = m.end() - 1  # position of the '{'
+        found = _find_brace_content(text, brace_start)
+        if found:
+            content, end = found
+            # Recursively expand any bracket macros inside the content
+            content = _expand_all_bracket_macros(content)
+            left, right = _BRACKET_MACROS[macro]
+            result.append(f"{left} {content} {right}")
+            pos = end
+        else:
+            # Unbalanced brace — emit macro text literally and continue
+            result.append(text[m.start() : m.end()])
+            pos = m.end()
     return "".join(result)
 
 
diff --git a/scripts/markdown_postprocessor.py b/scripts/markdown_postprocessor.py
index 92fae90ba..e3b99b928 100644
--- a/scripts/markdown_postprocessor.py
+++ b/scripts/markdown_postprocessor.py
@@ -470,11 +470,13 @@ def inject_field_metadata(text: str, idd_index: dict[str, IddObject]) -> str:
 
     Scans the markdown for heading patterns to determine the current IDF object,
     then looks up field metadata from the IDD index and inserts styled attribute
-    blocks after each field heading.
+    blocks after each field heading.  Also inserts ``<hr>`` separators between
+    consecutive IDF objects so long group pages are visually segmented.
     """
     lines = text.split("\n")
     result: list[str] = []
     current_object: IddObject | None = None
+    seen_first_object = False
 
     # Heading patterns
     # Object headings: # ObjectName or ## ObjectName (for group pages)
@@ -483,8 +485,6 @@ def inject_field_metadata(text: str, idd_index: dict[str, IddObject]) -> str:
     field_pattern = re.compile(r"^####\s+Field:\s*(.+)$")
 
     for line in lines:
-        result.append(line)
-
         # Check for object-level headings (h1, h2, h3) that match an IDD object
         h_match = h1_pattern.match(line)
         if h_match:
@@ -492,9 +492,18 @@ def inject_field_metadata(text: str, idd_index: dict[str, IddObject]) -> str:
             # Strip any Pandoc attributes like {#id .class}
             heading_text = re.sub(r"\s*\{[#.][^}]*\}", "", heading_text).strip()
             if heading_text in idd_index:
+                # Insert a visual separator between objects (skip the first)
+                if seen_first_object:
+                    result.append("")
+                    result.append('<hr class="idf-object-separator">')
+                    result.append("")
+                seen_first_object = True
                 current_object = idd_index[heading_text]
+            result.append(line)
             continue
 
+        result.append(line)
+
         # Check for field headings
         field_match = field_pattern.match(line)
         if field_match and current_object: