No; we coded the simulation in C++. Platforms like Arena are there, but I would suggest not to link them to C# or C++ to maintain speed and flexibility.
AnyLogic sounds to be interesting for a simulation when it is necessary to combine several approaches, for example, discrete event and agent-based modeling, with good opportunities for visualisation.
No simulation software, however sophisticated it is, can be compared with the unlimited possibilities when you just write a program to do that simulation exactly the way you want it to be done. That being said, the choice of programming language used is also crucial.
Personally, since this is all about number crunching, I believe that the best option is by far Fortran (version 95, 2003, or 2008). A popular choice is C/C++, but those languages are just not made for what you are trying to do. The primitive (essentially non-existant) matrix support in C, and the awful matrix support in C++ should be enough justification for rejecting them as the programming language of choice for anything involving Mathematics. However, i could add a long list of other reasons too. It is extremely easy to write a memory-leaking program in C/C++ (the language lets you do things with pointers that you shouldn't be able to do, and it's not even easy to figure out where the problem is after that). Furthermore, Fortran is module- or object-oriented; it's up to you which way you will go, and you can even mix them if you need to. So things that you can easily do in Fortran are often hard to do in C/C++, unless you use tricks to dodge the unecessarily complicated and badly designed core of the language. Not to mention the much better compiler optimization in Fortran. The list goes on, and it is a long one.
Honestly, I don't understand how people prefer to do simulations in C/C++. In any case, I shouldn't be surprised. Nowadays, many people even use... Java or Python or whatever similar, for writing mathematical applications. It sounds crazy, but it is true.