Ya know, if CL would put the detailed logging I described into the rendering code and make it available for testing to registered users, we might be able to put this whole memory leak thing to bed pretty quick. That's how me and my peers helped IBM solve a memory leak in one of their products. They created debug versions to dump logs, and we ran the code and collected the data for them. If CL is serious about solving the problem, then I am more than willing to run a debug version to help them find the problem.