When you approve something for launch, and then kill the product before it can either launch or launch effectively, then you become liable for the investment in the launched failure. Google can absolutely be sued by those that just had a failed launch. They misrepresented themselves. This risk, tends to be higher for the company, than allowing people to guess that the product is about to die, because game products tend to be rather large investments.
I can “launch” (pun intended) AAA title on a cloud VM in the time it takes for steam to download the game.
I haven’t used stadia; honestly curious.
There's stories like this [0] which suggests that the performance for Stadia was different than other platforms you might deploy for. The threading behaviour is a little bit different than just a VM, which is to be expected, but can come out surprising. So there's probably Stadia-specific patching for different games.
[0] "Stadia Adventures in slow server code on Unity" https://www.youtube.com/watch?v=s-SpWSEWYbU
I once casually speculated to a director that product X felt like it was going to get the axe. He was visibly freaked out and responded with a fervent Shakespearean lady-doth-protest-too-much denial, complete with demanding to know where I'd heard this from. Inevitably, it turns out that X's days were already numbered, but he already knew and I didn't.