At 3.2ghz, it won't hurt. Even vcore on auto won't matter on such a conservative clock. If you start aiming anywhere north of 3.6 I'd seriously reconsider this approach, not only are motherboards historically fond of overvolting, they're not always very good at getting the balance right. Sometimes voltages have to be decreased to find stability.
It's just about possible that this is no longer true, as with setter I haven't seen any particularly damning reports on x58 boards overvolting. I consider it unlikely that since P45 this has been perfected.
qpi does need to be within 0.5V of the ram voltage, so at 1.65V keep qpi at least 1.15V. A greater difference is considered to degrade the chip. Errors inherent in measurement and voltage control give the more conservative estimate of keeping them within 0.45V of each other. The main effect is to justify going over 1.65V on the ram. As most people end up with a qpi around 1.3V for 4ghz, and this allows 1.75V on the ram, and few people are willing to go that far over the intel guideline, it's something of a non-issue.
The 280gtx dying sucks, but it was probably unrelated to the processor overclock. A tenuous link is the power supply, if it is inappropriate for the rest of the system then the additional load from the processor may have been enough to drive it out of spec with a knock on effect on the card. This is unlikely. What power supply are you using? In what way did the card die?