I think these days that chart is mostly irrelevant. It is listing Core temps and most GPU algorithms these days are memory hard and memory temps really depends on memory type. the best source for safe temps is to read the datasheet for whatever memory manufacture / type your GPU is utilizing. But 90% of the time when a GPU fails its normally a capacitor or mosfet and is an easy fix. When im looking to buy broken GPUs the listings are almost always the lower end brands like Zotac because they use cheap capacitors and mosfets to hit there price target. The cheaper parts just don't hold up against the 24/7 usage.
Not really, I think there are nearly the same number of core and mem intensive algos.
I had 2 GPU´s running over months with 85°C, but they still run fine
but most of my gpu´s running between 40°C and 60°C