Strategy 3 is extremely interesting, from the "if you are repeating yourself, you are going wrong" point of view. Nevertheless such an approach needs a well-designed and complete "chapters tree" as step one; the best would be that titles of sections/subsections could be wrote one only to avoid any possible mistake.
A simple Excel page (or similar) would be probably the simplest way to produce the template code to be seeded into pages containing the start of chapters and ignoring other pages.
With very few simple conventions, anything - but pagelist - could be automatized, if a thorough seeding of such templates could be done as proofreading step 1, since a script could add too needed section tags, then build pages tags and fill chapter list into Index page, and all the needed code for ns0 pages/subpages.