So it appears that the Cary tech was correct. Your readings were wrong. I have to say, due to poor communication with the manufacturer you end up proclaiming that you'd never buy a Cary product again if your life depended on it despite the issue was user error as Cary correctly pointed out. Ugh.
I didn't even think that anyone would consider setting bias with music playing. Yes, you want your speakers connected when you check bias but DO NOT play music while measuring. The whole point is that you're checking the bias while the amp is at idle, and running a signal through the amp is not idle. Depending on how you had the volume knob set you could have easily been running the amp 50mA or more ABOVE idle when you were measuring and hence frying your power tubes.
I didn't even think that anyone would consider setting bias with music playing. Yes, you want your speakers connected when you check bias but DO NOT play music while measuring. The whole point is that you're checking the bias while the amp is at idle, and running a signal through the amp is not idle. Depending on how you had the volume knob set you could have easily been running the amp 50mA or more ABOVE idle when you were measuring and hence frying your power tubes.