Thanks for sharing this Baker.
I'd like to understand what's going on better though. I can reproduce it using only winquake 1.09 (record demo, map e1m1, noclip to the exit, then play the demo. intermission camera is pointing at the wrong angle), so it looks like a bug in the original code. Weirdly, one recording attempt produced a demo that plays back correctly (every time I play that particular demo file). So there appears to be some randomness to it at recording time.