Another try to fix lists

            # Insert page break before "Scope" Heading1 and "History" Heading1

            if re.match(r'^1\s+Scope\s*', full_text):

                insert_page_break_before_heading(para, full_text)

        # Annex heading: "Annex A: Title" or "Annex A (informative): Title"

        elif (match := re.match(r'^(Annex\s[A-Z](?:\s*\((?:informative|normative|informative or normative|normative or informative)\))?:)\s*(.+)$', full_text, re.IGNORECASE)) is not None:

        # Annex heading: "Annex A: Title" or "Annex A (informative): Title" or "Annex: Title"

        elif (match := re.match(r'^(Annex\s[A-Z](?:\s*\((?:informative|normative|informative or normative|normative or informative)\))?:|Annex:\s*)\s*(.+)$', full_text, re.IGNORECASE)) is not None:

            annex_letter = match.group(1)

            annex_title = match.group(2)

            old_id = abstract.get(qn('w:val'))

            abstract.set(qn('w:val'), abstract_id_bn)

    def split_run_with_breaks(para, run):

        import copy

        try:

            index = list(para).index(run)

        except ValueError:

            return

        elements = list(run)

        new_runs = []

        current_run_elements = []

        run_props = run.find(f".//{{{ns['w']}}}rPr")

        def create_run(elems):

            r = OxmlElement('w:r')

            if run_props is not None:

                r.append(copy.deepcopy(run_props))

            for el in elems:

                r.append(el)

            return r

        for el in elements:

            if el.tag == f"{{{ns['w']}}}rPr":

                continue

            if el.tag in (f"{{{ns['w']}}}br", f"{{{ns['w']}}}cr"):

                if current_run_elements:

                    new_runs.append(create_run(current_run_elements))

                    current_run_elements = []

                new_runs.append(create_run([el]))

            else:

                current_run_elements.append(el)

        if current_run_elements:

            new_runs.append(create_run(current_run_elements))

        para.remove(run)

        for i, r in enumerate(new_runs):

            para.insert(index + i, r)

    for para in paragraphs:

        # Skip if already processed

        if id(para) in processed_paras:

                counter_compact += 1

                continue

        # Pre-process: split runs with breaks

        for child in list(para):

            if child.tag == f"{{{ns['w']}}}r":

                 if any(node.tag in (f"{{{ns['w']}}}br", f"{{{ns['w']}}}cr") for node in child):

                      split_run_with_breaks(para, child)

        # Get all direct child runs (not nested runs)

        runs = para.xpath('./w:r', namespaces=ns)

        if not runs:

            if child.tag == f"{{{ns['w']}}}r":

                text_elem = child.find('.//w:t', namespaces=ns)

                if is_start_of_line and text_elem is not None and text_elem.text:

                if text_elem is not None and text_elem.text:

                    text_stripped = text_elem.text.lstrip()

                    if text_stripped.startswith('- ') or re.match(r'^\d+\.\s', text_stripped):

                    is_numbered_candidate = bool(re.match(r'^\d+\.\s', text_stripped))

                    is_bullet_candidate = text_stripped.startswith('- ')

                    if is_numbered_candidate or (is_bullet_candidate and is_start_of_line):

                        list_item_runs.append((idx, child, text_elem))

                # Update state