think the most solid way, and the way I'd do it now if starting my PhD all over, would be a bunch of org docs with LaTeX weaved in.

Pandoc I think really e a different niche, hard to imagine really a complicated document benefiting from even more pipelining. You're editing a lot...