Thank you very much for running those tests and providing all the information, those really are very useful to us. From my initial analysis it does look like the process is holding on to some objects which are larger than the expected values would be, there are also a number of objects which look like they could have been cleaned up and memory freed. I will be carrying out more extensive testing this week to see if I can perform any refactoring to improve the performance of the conversation process. I will keep this thread updated with any improvements that I am able to make.
David Gault
