Имея в виду Комментарии Mr.E. и Arran я сделал свой список полностью на CSS селекторах. Самое сложное было с моей собственной структурой списка и отметками (смена классов и т. д.), а также о создании необходимых селекторов на лету и сохранении их в памяти во время обхода.

Я избавился от ожидания нескольких элементов путем поиска всего, что не является состоянием загрузки. Вы также можете использовать селектор ":nth-child", как здесь:

#in for loop with enumerate for i    
selector.append(' > li:nth-child(%i)' % (i + 1))  # identify child <li> by its order pos

Это мой жестко прокомментированное кодовое решение, например:

def parse_crippled_shifted_list(driver, frame, selector, level=1, parent_id=0, path=None):
    """
    Traversal of html list of special structure (you can't know if element has sub list unless you enter it).
    Supports start from remembered list element.

    Nested lists have classes "closed" and "last closed" when closed and "open" and "last open" when opened (on <li>).
    Elements themselves have classes "leaf" and "last leaf" in both cases.
    Nested lists situate in <li> element as <ul> list. Each <ul> appears after clicking <a> in each <li>.
    If you click <a> of leaf, page in another frame will load.

    driver - WebDriver; frame - frame of the list; selector - selector to current list (<ul>);
    level - level of depth, just for console output formatting, parent_id - id of parent category (in DB),
    path - remained path in categories (ORM objects) to target category to start with.
    """

    # Add current level list elements
    # This method selects all but loading. Just what is needed to exclude.
    selector.append(' > li > a:not([class=loading])')

    # Wait for child list to load
    try:
        query = WebDriverWait(driver, WAIT_LONG_TIME).until(
            EC.presence_of_all_elements_located((By.CSS_SELECTOR, ''.join(selector))))

    except TimeoutException:
        print "%s timed out" % ''.join(selector)

    else:
        # List is loaded
        del selector[-1]  # selector correction: delete last part aimed to get loaded content
        selector.append(' > li')

        children = driver.find_elements_by_css_selector(''.join(selector))  # fetch list elements

        # Walk the whole list
        for i, child in enumerate(children):

            del selector[-1]  # delete non-unique li tag selector
            if selector[-1] != ' > ul' and selector[-1] != 'ul.ltr':
                del selector[-1]

            selector.append(' > li:nth-child(%i)' % (i + 1))  # identify child <li> by its order pos
            selector.append(' > a')  # add 'li > a' reference to click

            child_link = driver.find_element_by_css_selector(''.join(selector))

            # If we parse freely further (no need to start from remembered position)
            if not path:
                # Open child
                try:
                    double_click(driver, child_link)
                except InvalidElementStateException:
                        print "\n\nERROR\n", InvalidElementStateException.message(), '\n\n'
                else:
                    # Determine its type
                    del selector[-1]  # delete changed and already useless link reference
                    # If <li> is category, it would have <ul> as child now and class="open"
                    # Check by class is priority, because <li> exists for sure.
                    current_li = driver.find_element_by_css_selector(''.join(selector))

                    # Category case - BRANCH
                    if current_li.get_attribute('class') == 'open' or current_li.get_attribute('class') == 'last open':
                        new_parent_id = process_category_case(child_link, parent_id, level)  # add category to DB
                        selector.append(' > ul')  # forward to nested list
                        # Wait for nested list to load
                        try:
                            query = WebDriverWait(driver, WAIT_LONG_TIME).until(
                                EC.presence_of_all_elements_located((By.CSS_SELECTOR, ''.join(selector))))

                        except TimeoutException:
                            print "\t" * level,  "%s timed out (%i secs). Failed to load nested list." %\
                                                 ''.join(selector), WAIT_LONG_TIME
                        # Parse nested list
                        else:
                            parse_crippled_shifted_list(driver, frame, selector, level + 1, new_parent_id)

                    # Page case - LEAF
                    elif current_li.get_attribute('class') == 'leaf' or current_li.get_attribute('class') == 'last leaf':
                        process_page_case(driver, child_link, level)
                    else:
                        raise Exception('Damn! Alien class: %s' % current_li.get_attribute('class'))

            # If it's required to continue from specified category
            else:
                # Check if it's required category
                if child_link.text == path[0].name:
                    # Open required category
                    try:
                        double_click(driver, child_link)

                    except InvalidElementStateException:
                            print "\n\nERROR\n", InvalidElementStateException.msg, '\n\n'

                    else:
                        # This element of list must be always category (have nested list)
                        del selector[-1]  # delete changed and already useless link reference
                        # If <li> is category, it would have <ul> as child now and class="open"
                        # Check by class is priority, because <li> exists for sure.
                        current_li = driver.find_element_by_css_selector(''.join(selector))

                        # Category case - BRANCH
                        if current_li.get_attribute('class') == 'open' or current_li.get_attribute('class') == 'last open':
                            selector.append(' > ul')  # forward to nested list
                            # Wait for nested list to load
                            try:
                                query = WebDriverWait(driver, WAIT_LONG_TIME).until(
                                    EC.presence_of_all_elements_located((By.CSS_SELECTOR, ''.join(selector))))

                            except TimeoutException:
                                print "\t" * level, "%s timed out (%i secs). Failed to load nested list." %\
                                                     ''.join(selector), WAIT_LONG_TIME
                            # Process this nested list
                            else:
                                last = path.pop(0)
                                if len(path) > 0:  # If more to parse
                                    print "\t" * level, "Going deeper to: %s" % ''.join(selector)
                                    parse_crippled_shifted_list(driver, frame, selector, level + 1,
                                                                parent_id=last.id, path=path)
                                else:  # Current is required
                                    print "\t" * level,  "Returning target category: ", ''.join(selector)
                                    path = None
                                    parse_crippled_shifted_list(driver, frame, selector, level + 1, last.id, path=None)

                        # Page case - LEAF
                        elif current_li.get_attribute('class') == 'leaf':
                            pass
                else:
                    print "dummy"

        del selector[-2:]