Well, I accidentally discovered what is causing the crash while doing testing. If i client that requests a transcode spams the server with stream start/stops (by hitting play/stop quickly) after 5 or 6 times it will cause the segfault. It is repeatable every time but the exact number can vary from 4 to 8 or so.
Not sure why this is causing the driver to crash but i suspect its the way the transcoder is starting and stopping sessions the GPU gets confused as to what it’s supposed to be doing. Need to find a way to rate limit that…
