I am actually using this in a very important product. Although it hasn't been into full production yet, so far in qa testing it works very well.
In terms of volume expression eval works best when resulting functions are cached whenever possible for speed. Remember that compilation is still ebing done so if you are compiling an expression that you have lready compiled before you are introducing
Otherwise I see no reason for volume to affect expression eval output.
The risk I guess is that there may be a few implementation details that might not work as expected. This is because the c# implementation isn't complete yet
But if it runs your expression and gives you the desired output then I'm sure it will perform its job in a production environment.
Let me know if you encounter any problems.