Often this is a conundrum of naive technologists clamoring to shove untested tools into production for... reasons? Excitement? Risk-taking adrenaline? I'm not quite sure.
Yes Google's marketing oversells (or even lies a la Google Stadia), but at the end of the day I think Google SREs/devs/evangelists expect (wrongly) some measure of common sense and restraint from users and that is just not there. For example, if you ever watched some of Kelsey Hightower's earlier kubernetes talks he practically begs users to deeply consider their use case and whether or not they should actually be using kubernetes.
Google also expects their users to SRE/devops/refactor/upgrade constantly like they're able to. Which is also insanely unrealistic.
Ultimately, while some cases are Google's fault, often the blame is shared - most of the time. The market realizes this and thus the opportunity cost is already priced in somehow.