I think the biggest difference here is that GPUs are fairly general-purpose devices. Even if you only need one as a cheapest possible way to output something to your monitor, it's easy to imagine how that might change and why you might as well future-proof your setup.
The other thing you didn't mention about why a 1030 is manufactured e-waste is in large part the competition from the used market. You can get equivalent functionality for a few dozen dollars, so buying a brand new 1030 as an individual makes no sense. This isn't a problem in the specialty radiation device market.
The thing is that most people probably don't buy Geiger counter modules or devices because of a burning need to do some specific tasks, like you would with GPUs. I'll bet that most people buying the Adafruit kit are getting it as a novelty, or for educational purposes, or to just have something reporting the background radiation around them out of curiosity. Because of this, price is the most important factor for these groups.
I've never heard of Radiacode, and looking it up, it seems seriously impressive - especially the portable gamma spectrometry that you mentioned. But I also know that it's a very niche device, and most people who would shell out $250+ for this need it for chemistry projects, exploring irradiated areas, evaluating radioactive items and so on. It's not a lot of people, considering that if you just want something to give you a readout, you'll get the <$99 product, and if you need it occupationally, you'll already have received a professional device.
I had someone tell me a while back $100 is an adult $20. As in, what you saw as a kid as $20 of fun but factoring in how long it took you to get the $20, it's about the same. I know that math will be massively variable, but I like the idea of equating kid brain money to adult money for "toys" like this. A $5 toy vs a $40 toy, as a kid, is a big deal. A $40 toy vs a $70 toy almost isn't, because they're both already "expensive".
I suspect for most on HN, their toy-money equivalence leans to the Radiacode or equivalent being the better buy.
For me, a $100 purchase is something I can probably justify if I think I'll enjoy it a lot. $250 is still a 2.5x upcharge, and I consider these expenditures for a long time, making sure that I really actually need it.
And then, the $100 is just an example reference. It's sold by Adafruit, whose selling point is trading some value in exchange for reliability, convenience and compatibility. When you look at the market in general, I can find countless cheap pre-built Geiger counters at prices that are a lot closer to $50. I can't vouch for their accuracy, but there are so many unique options that I imagine that at least a few will be pretty decent.
I do like my Flipper for how adaptable it is but then end up using it to copy RFID tags as primary purpose.
https://www.aliexpress.com/item/1005004551394154.html this is 32eur
+ ~5eur for an esp32 board and some wires
+ http://docs.espgeiger.com/output/webportal
This gives you a measurement node, with mqtt support, online data export and a web interface.
For ~40eur all together, it's still very cheap.
Meanwhile, Raysid or Radiacode let you do full on gamma spectrometry, for about 2.5x the price.
And, like, yeah, 2.5x the price. I'm not oblivious to that being significant, but to me it's sorta like an GT1030 with DDR4 - it's about $100 or you could get an RTX 3060 for a bit over $300. Like, you might have a use where you ONLY need a cheap display card, but there's a reason LTT calls the 1030 eWaste from the factory. Similarly, you might get enough out of the Geiger or it might do something the Radiacode doesn't for you but for most hobbyist, I doubt it.