DirectX, OpenGL, and Vulkan provide common API for developers to use, and provide a layer between software and the hardware. That's why you see admin request for software like afterburner, which will directly access, it needs to bypass the layer and use driver/hardware directly. These APIs rely on the manufacturer drivers.
Software can be written to stress hardware like how prime95 uses certain instruction sets and you need to have proper cooling to handle them. GPUs are quite complex with many features (AA, ray tracing, occlusion, shaders, lighting, etc...). Now you can imagine that more features and computations done, along with the rate (FPS) all drive usage, heat, and power.
Now the one GPU that was getting "destroyed" in the past, supposedly by New world, was really, "...EVGA revealed that they discovered, via X-ray analysis, that it was instead “poor workmanship” that killed the cards,..." I perused the links you provided, and the reports said it is quite small, and gigabyte seemed to be the most common factor for some.
As long as a game utilizes the API layer, it is highly unlikely to kill a GPU as long as it is properly cooled and running at normal power and settings. Even then, today CPUs and GPUs have thermal throttling safeguards. Games can be more considerate for heat and power concerns by not running unconstrained on menus (3000 FPS on menu).
