Somehow I don't think this surprised anyone. I just did a quick check and the same people who were making these kinds of decisions when I was there all appear to still be there (at least by their social status indicators like LinkedIn, Twitter, and Facebook). Back in early 2009 when "the mortgage crisis" was the big financial scare, I talked with folks about the difference between "moon shots" and "products".
The former (also called Black Swan events in Finance) are things that change everything all at once, the latter are things that a company produces to diversify their revenue risk. Google had just released the Dream (the G1 phone) and given all the employees one so we could "eat our own dogfood" and the biggest takeaway was that the biggest issue here was that this sort of "product" really really needed a real customer support function. Google was fixated on using "smart people and technology to completely eliminate the need for 'lame' customer support that was expensive and needed managers and stuff." One 'senior' engineer (who was 'senior' by virtue of their time at Google, their first job out of school, and not because of their years of industry experience, explained to me that by 2011, 2012 at the latest, no company anywhere would use "people" for customer support, it would all be web driven, "knowledge trees", and smart dialog agents that could do anything real people could do.
I had worked previously at NetApp where they spent a ton of money and invested a lot in insuring their customers were supported, and I tried to explain that it wasn't the '90%' problem it was a '10%' problem. Your automated system can deal with 90% of the problems, but 10% of them were 'weird', special cases, that no automation would be able to deal with. And ignoring that, and leaving 10% of your customer base angry and unsupported, was death for any retail product.
I see the same thing with 'self driving' too, your car can do 90% of the things you need it to do on its own, and then that last 10% are the things that kill you.
When Stadia was announced I looked to see if they had learned about the need for real people support and it didn't appear they had. And they were dealing with "the Internet" which for engineers looks like a smooth low latency pipe between your computer and theirs, but for real people is a series of networks that have to be transited with support levels that vary between enterprise level and sixth grader enthusiast level. And with all that randomness between the product and the customer, no 'people' support to create a good "we hear you, we're on it" experience and more importantly a strong feedback loop to engineering/marketing/sales about what is causing the most trouble for prioritization of resources, and poof. It was dead on arrival. It just took a while for Google to figure that out.
I mentioned in my exit interview that I hoped they would internalize that problem before they ran out of search advertising revenue runway, but 10+ years on it looks like that isn't happening.