Very large ETH miner here, we use GPUs. Mostly 4-5 year old tech too. Unlike BTC, ETH does not require the latest hardware because the algo, ethash, is memory bound. This is a great explainer [1].
You may be implying that the majority of the network is custom ASICs. I doubt it is more than about 30%, if that. Come august, all the ones with 5gb of ram will also disappear off the network because of DAG growth.
GPUs are also not destroyed by using them. I've got GPUs sitting in the worst possible conditions, running just fine for many years now. We have started to change the thermal paste on the hot ones though... it dries out. I don't know what you were dealing with, but that sounds like something else caused the failures as the chips themselves don't wear out like that.
[1] https://www.vijaypradeep.com/blog/2017-04-28-ethereums-memor...
Fans are moving parts and require electricity.
We also put 12 cards in a case instead of the usual 8 or 9.
This means fewer other parts too... psus, mobos, ram, ssd, case fans... all of which use power.
In fact, smart miners overclock/undervolt their gpus... so they are running faster, but with less power. Power is what creates heat and heat is what hurts things. Except it really doesn't.
Fans certainly die out though... we go through a lot of those on our cases. PSUs are also generally failure prone since they tend to be hand soldered on the insides.
This is such an urban legend. There's been quite a few tech people testing this and there's countless articles on the topic. In short - I have not seen a single one which actually found real significant issues with the hardware, even if some are listed in theory. If this was common at least one of them would buy a severely degraded batch, or we'd see some proven report about this.