Sounds like really bad coding taken to another level. At some point in your program you will know which thumbnails are needed for the page. Instead of structuring that logic and commenting it well, perhaps even turning it into a sub-system or library, you pepper the templates with database calls. And then, you waste more CPU time by scanning the full HTML output of a page multiple times.<p>This is like AJAX all on the server side. All the hassle of async processing with none of the benefits.
Wow. This is a really ugly workaround for a problem that began when you started using MVC. I would assume that any savings in db calls is completely wasted in processing overhead. Even more was wasted just coming up with this rigmarole. More reasons why MVC is bad juju.
"Datex"? Scary. Not sure I agree that the example given at the beginning of the article is a big problem (why not just make the model apis more fine-grained?), and also not sure what this has to do with MVC specifically.
Do you have any stats on the CPU/memory usage difference using this approach? Obviously your database/memcache calls are optimized, but I'm curious what the overall effect is.