I would love to see proof of said "rule of thumb" actually being true in real life.
I don't have stats for GPU but back when overclocking CPU was the craze, there were quite a few users who had their highly overclocked and hot CPU die after a few months. If you dig up electronics and semicon datasheets, there are manufacturer charts for expected lifespan vs temperature delta.
IMO, heat is way overblown. GPU's are safe up to 90c EASILY, and I've only ever heard of a GPU dying from heat over 110c. Keep it below that, and in all likelyhood your card will be absolutely fine.
Which is why I said it might not matter if you intend to replace the GPU with a faster one as soon as they are available. The manufacturers usually set the shutdown/throttle point at some temperature they expect is the max allowable that doesn't degrade the expected lifespan near their warranty timespan.
However, the risk is that if you push it to that limit, the delay between thermal protections kicking in could still kill your GPU within a few months like those dead CPUs.
So while I think it's silly to spend money on massive cooling equipment to keep temperatures as low as possible, I also don't think it's good to run the GPU right up to the throttling/shutdown point. 10C is a healthy buffer so I try to keep things to 80C max.