Not sure I'd go as far as "single largest central of survivorship bias on the internet" but I understand the sentiment. As another commenter noted used RTX 3xx

Not sure I'd go as far as "single largest central of survivorship bias on the internet" but I understand the sentiment. As another commenter noted used RTX 3xxx series cards are at most two years old. Given the anticipated lifetime of a card as the Bathtub Curve applies I'd argue any surviving used card is pretty firmly in "escaped infant mortality" territory. This applies to gamer cards as well (of course) but this is where my other points apply. Anecdotal but in 2017 I had 200 GTX 1xxx cards for a personal mining project (it was interesting). From what I remember I experienced <5 card failures within the first few months but smooth sailing otherwise. I kept a few for personal use and eventually sold the rest with a handful of them (small sample size, I know) going to personal friends. They're all still running strong. For the remainder that were sold on eBay I only had one return (it was fine, of course) and all of the reviews were impeccable - along the lines of "I know this card was used to mine but I'd swear it's new 10/10". Note that I was very clear and transparent that they were used for mining in the item description and I even included pictures of the mining environment in the listings. My eBay account is still active and I haven't heard anything from a single customer over the past three years (I assume they'd come back in the event of a card failure). I think 200 cards is a large enough sample size to bolster my argument. As another commenter also noted many of them are being sold as new. This is typical crypto ecosystem shadiness but for a miner to be able to even pass a card off as new speaks to the high level of care higher-volume professional mining operations employ. For modern Nvidia GPUs (even going back to Pascal) I'd argue anything used for less than two years is essentially burn-in testing.