I know like 4 big Google projects and all seem to be failures (Kubernetes being the most successful, but still in hype so nobody realizes yet how mediocre it is). Right now it seems a good time to start working on a Google killer.
You can easily see that a solution is mediocre if you see that it doesn't attempt to solve any of the hard problems (in this case networking, network debugging, storage, performance) while requiring lots of overhead to do a trivial thing (e.g. here are my static html files, no go and host these).
It works okay on AWS/GoogleCloud but at least on AWS you already have all the features of k8s on IaaS level. On everything else it runs shitty and you spend more human resources on getting it to work / keeping it working than when you do the scheduling yourself.
> Right now it seems a good time to start working on a Google killer.
If all Google had was the handful of failures you claim to be (rather selectively) aware of, there'd be no need for a Google killer.