> Ultimately XHMTL is better suited to epub's use case, which is static documents that are written once and not dynamically generated by templating languages or libraries.

That's backwards. Relying on software to assign random meaning to broken input is never a good solution to anything, neither for static documents nor for dynamically generated documents. It's just that HTML allows you to get away with it, kindof (i.e., no scary error messages, just subtle breakage everywhere), which is why people built templating systems that didn't bother with correctness either. If the web had been XHTML from the beginning, templating systems would have been built in a way that makes it hard to generate broken documents in the first place, because the error messages would have created the incentive to do so, thus avoiding a lot of the subtle breakage as well.