intelligently detect header type elements without depending on the std_header selector?

Try this:

repeat for y = 10, 20, 30, 40, 50

• find the element at point (page_width/2, y), using document.elementFromPoint()
• if it's different from all elements found so far, check if it has any ancestor which has position:fixed. check if such ancestor has both width and height > 100 (say) and lies within the top 300 pixels of the screen. if so, consider it a 'fixed header' type element for calculation of unusable space at the top.
• if more than one fixed-header-type element is found, consider the bottom most one

(Even if this works, the std_header selector should still be part of urlData since it's useful semantic data anyway?)