While working on the DDT project, I came to the same conclusion. DLTK was very promising and served us (the DDT project) well in the beginning, but eventually a lot of roadblocks were stumbled upon. There were several minor API changes we needed for DDT, in order to move things forward, and those were not being addressed. And those were the simple ones, never-mind bigger API or functionality changes. By the time of DLTK 4 and 5, I could see its development slowing down.
To address this, I started forking and rewriting bits and pieces of DLTK. Because these pieces were mostly UI components, we could used them on top of a DLTK-based IDE, by just replacing some DLTK components as needed (for example, certain preference pages, editor functionality, wizards, etc., one by one). Some were written mostly from scratch, other were minor refactorings of existing DLTK/JDT code (since the code is EPL, it's not a problem).
As the amount of refactored code grew, I turned it into a separate toolkit that could be used by other IDEs, even non-DLTK based ones. The motivation was me being able to use that code in the Goclipse IDE, which is an IDE project I recently started contributing to as well. (GoClipse is not DLTK based though, nor did I want to turn it that way) This "toolkit" code is available here:
https://github.com/bruno-medeiros/LangEclipseIDENote that the functionality available there is a very, very small fraction of what DLTK offers (although the little that is available I do believe is written in a much more clean way that most DLTK/JDT code). I do hope to add more functionality to it in the future, but I'm just a one-man show working part-time on this, so don't expected anything big anytime soon. Contributions welcome ;) , if you think it could be of use to yourselves.