Is it safe to use different versions of cuda runtime within a single process?

Hello everyone! I have a large project with many modules in it and some of them are using CUDA. Not a long time ago we started getting crashes in our app and right now I am trying to identify what could be the reason behind this. I have noticed that we have 2 modules that are used from the same Windows process. One of them uses Optix 3.7 that is build with CUDA 6.5 and another one includes our GPU code and it’s built with CUDA 7.5.

Is it a safe practice to have code that uses different versions of CUDA runtime within a single working process? Does anyone here have any experience with such a configuration?

Thanks in advancce!