I'm sort of shocked that you're shocked. The rigid and inflexible architecture of systems like Rails and Django are the primary reasons for WP's success. There's a reason WP has more themes and plugins than these other systems, and that reason is ease-of-programming.
Every time I see some programmers advocating the use of highly rigid architectures or programming models, I'm forced to remind myself of Lisp. Lisp is the programming language. Lisp created most of the fundamental constructs you see in modern programming languages today. It's quite possible to do damn near anything in it, and it's arguably as close to the perfect mathematical construct of a computer language that can be made. It's a bit rigid and weird in it's syntax, but the list model is just great once you wrap your head around it.
On the other hand, the language that is clearly the most successful one ever is C, and it has often been argued to be the worst language ever. It has arcane syntax, it requires detailed knowledge of memory pointers and other oddities, and it's very much a lower level language. And yet it is clearly the most successful language ever. Why? Not because of its model or architecture (which are frankly terrible), but because of its flexibility. Programmers both inexperienced and experienced can use it to produce powerful systems. It covers the whole gamut of skill levels.
That's the problem with rigid architectures. Programmers tend to flock to them as they become better at their craft, because they provide organization and a system of management of the large scale code they're creating. But inexperienced programmers find the learning curve just a bit too steep.
You can't go from "Hello World" to creating a new WP theme in an MVC style. But you can go from basic HTML and PHP knowledge to creating a theme modification or a plugin. It may not be the technically best solution, but by god it'll get the job done for now. Speed of delivery matters, and if the choice is between doing-it-right and doing-it-right-now when you have a deadline and need to get paid, guess which path is going to get chosen?
Code has a limited life expectancy. It doesn't survive long in the wild, and this time is more and more frequently measured in years, not decades. So it doesn't make a lot of sense, in the long run, to spend extra time to build around a large framework when, damnit, we've got work to do. Every moment you spend learning yet-another-framework is a moment you didn't spend creating the next product and earning the next dollar.
Systems that are fast to learn and flexible enough to get the job done quickly always win over systems that are rigid and hard-to-learn.