I think implicit in this myopia is they don't eat their own dogfood. If they have their best people managing their core products running their own infrastructure which is clearly much more reliable than GCP, then GCP won't get better. On top of which their best people will move toward their own internal services.
Contrast this with the other cloud providers who seem to be using their own clouds for their primary services creating a feedback loop which ensures they don't have the big outages that GCP has had over the past year.
Dogfooding usually helps with usability. I'm not sure it has much to do with big outages; GCP is big enough that all the individual decisionmakers are pretty maximally incentivized not to break it.
Contrast this with the other cloud providers who seem to be using their own clouds for their primary services creating a feedback loop which ensures they don't have the big outages that GCP has had over the past year.