Software rewrites are extremely challenging, and more often than not, fail (good example: http://www.joelonsoftware.com/articles/fog0000000069.html). Perhaps a more measured factoring approach over time would be more appropriate?
I don't understand the point of what he's proposing. Rewriting WordPress as completely as he wants and breaking all existing plugins and themes would surely just result in someone forking the existing code and maintaining a backwards-compatible version.
The ecosystem is the valuable thing, and they'd be insane to throw that away. All the problems he notes are basically just a result of WordPress growing and changing over the decade it's been around, and/or direct results of having originally been written in PHP 4.
Gradual refactoring, as you say, makes more sense. Otherwise you might as well just build an entirely new system from scratch.
WordPress is already going through a measured factoring approach.
Themes were ported over from a mishmash of procedural calls into the WP_Theme object in version 3.4. Posts (and pages, custom post types, etc) were ported over from an associative array and various procedural calls into the WP_Post object in 3.5.
But WordPress does - and likely always will - work to retain backwards compatibility so that existing themes and plugins won't choke after a core update. So even when newer APIs and refactored objects exist, its up to individual developers to 1) learn about them and 2) start using them.
