Maybe I am less tolerant about this stuff than most, since I have some e-sports background from fighting games (Virtua fighter specifically) from about 10-20 years ago. In VF, ping/latency/lag bigger than 16ms (1/60th of a second) had an observable effect on gameplay. Current modern day "standard" ping of 40-60 ms would've been barely tolerable or unplayable. Because of this we used to play on cathode ray TVs for a long time since most flat TVs had too much image-enhancing garbage that caused lag. Online gameplay was a joke back then.
I remember playing CS back when it was a Half-life 1 mod in beta phase with 200+ ms ping times. There were all sorts of discrepancies across players for latency. People just sort of dealt with it. Games adopted, expanded and improved mechanisms to account for those discrepancies.
I think the type of latency is as relevant as the actual numbers. By this I mean if you play a competitive online shooter game you have a monitor, device running the game, cable connecting the two, input devices, etc. None of those steps in the path are going to communicate instantly. This is not to say those communications cannot be "fast". Your inputs might be fast but the result of those inputs still needs to register on the server. This would be where those earlier mentioned mechanisms would apply. Lag compensation, netcode, whatever you wish to call it.
The underlying point is/was removing latency or "lag" completely isn't realistic. It never will be. This does not mean there is no way to compensate for it. Where I get concerned is, intuitively, Stadia would push all of that latency onto the inputs and outputs, including the way the video stream plays into them (conversely, the client/server agreement problems would presumably cease to exist, for the most part). I doubt those earlier mentioned compensation mechanisms can be formulated in the same way under those conditions. I'd be curious how they intend to compensate for it, if at all.
Lastly, I wouldn't think people looking for a competitive game experience are going to be blindly jumping on the Stadia bandwagon. Likewise, people focused on games where input lag is a huge deal likely aren't either. If they are they might want to rethink that approach. At least prior to testing out the service to get a read on how it behaves for their specific environment. In cases where neither of these two things are the desired experience, and the connectivity to the data center is excellent, the "it will be too laggy" comments feel premature and over-exaggerated.
Their best pitch is that it will have input lag of 166ms. Which even with my old slow 60hz monitors sounds horrible.
Assuming those are accurate numbers. I'd approach those type of numbers with a high level of skepticism without knowing the exact environment they were tested under. I wouldn't put it past a lot of these "tech sites" to fudge the numbers for reasons (or Google, for that matter). I wouldn't put it past them to make poor assumptions or base results on things they're unaware of as well. Even if those numbers were accurate all it says is they were accurate in that particular environment. The reality is the latency is going to depend on a long list of factors.