I'm not sure if this qualifies as an answer to your question since I'm not a professional (or researcher) in the area, but I have programmed in a couple of constraint programming systems. My object has been more about learning constraint programming and the systems than to solve a specific problem.
The problems has been very diverse: there is a lot of recreational mathematics and puzzles, but also traditional constraint programming problems, such as combinatorial problems as well as (decompositions) of global constraints.
The systems I have used so far are:
- MiniZinc
- Choco (Java)
- JaCoP (Java)
- Comet
- Gecode (C++)
- Gecode/R (Ruby)
- ECLiPSE (Prolog)
- SICStus Prolog
- Tailor/Essence'
For every system I start to learn, I begin with about 17 problems, so for each system there is at least these problems. However, for some systems there is a lot more models e.g. MiniZinc (over 600 models, small and large), Comet, ECLiPSe, and SICStus.
See www.hakank.org/constraint_programming/ for links to the models I've implemented in each system. Also, I blog about constraint programming at "My Constraint Programming Blog" (http://www.hakank.org/constraint_programming_blog/ )