I understood you were talking about multiple people. You seem to define an office environment as a realistic environment where people will use cloud gaming services and someone at his house reporting a good experience is somehow the best possible condition? What we should all be asking about every time someone reports a good or bad experience, are questions like what is your ping to the server? were you using a wifi or wireless connection? Chromecast Ultra or PC? office environment or your house? and other similar questions. These are the types of questions that leads to the truth and something I will try asking more myself to get a better idea.
This isn't the problem. I'm a network engineer. If you sit at home, you're not alone on the transportation network. If you're in an office, you're sharing a bigger network, but network level load isn't the issue here. The transportation network is the core of what Nostremitus is trying to get at. If you're in an office, you might share a bigger level two network with more people, but you're still sharing the level three network with as many people at home or at an office. You have different use cases in an office and at home, of course, which could theoretically change the experience, but let's look aside from that, at the transportation network.
On the transportation network on the router level, you'll have either BGP or OSPF routing protocols to find routes for your traffic. Without getting down into everything, even pretending you're tied directly to your ISP's router with no one else connected to the same router (and mind you, a typical router normally hosts thousands of users), you still have these routing protocols that are going to take care of traffic.
Those protocols ensure that your traffic is handled pretty fairly. The same protocol is the reason why TCP/IP have sequence numbers: Your route will vary. A lot.
Now, If you're on Google Fiber, they might incorporate headers that mark their traffic in a way that you get a prioritized static and short route. But probably in most, if not all, use cases, you're not in control of what goes on when your traffic goes out by itself on the great internet. These factors will contribute to varying ping. I don't know how Stadia handles compression, but compression is normally based off of a complete image, so it might be that if you're missing too many packets in a image, you can't decode it, so you'll have to wait for the whole image to arrive. That might make things variable.
So even if you compare this all with a somewhat laggy TV - and I can remember the time when I had a TV with 100ms input lag - you'd perhaps start to think Stadia would work well if you have a low latency screen, for the normal consumer. I don't think so, and I think a big part of that is the variable response time. Frame pacing, and even meta-frame pacing, like pacing of several frames, will be much more variable on Stadia. That's why one button press may have 100ms, while the next may have 130ms. Perhaps even 200ms in less optimal situations.
It's just one of those factors why I can't get excited about this platform.