I've been doing end user computing for quite a while, and we've gone through so many cycles of "where the client intelligence lives" or "where the virtual desktop is hosted" and everyone oscillates between two extremes. PCs to zero clients usually ends up being a mix of laptops and thin clients in the end. All VDI ends up being some VDI after some very expensive POCs in most cases. I guess the same debate of "host it yourself vs. rely on a cloud provider" is alive and well here. I see it every day where I work -- the management is all about cloud, and the staff are fine with some cloud, but going all the way over to total dependence on a third party is not great in my mind.
Something as fundamental as DNS should probably at least have some footprint in your "locus of control." I didn't say "in your office" but fundamental stuff that could completely kill everything else if you lost it shouldn't be given over to a third party that you don't directly control. In this case, Dyn had a DDoS attack, and on-premises DNS could too. But having a way to run both off and on premises makes good sense...if one entity is having a bad day, the other could at least keep things alive. However, all this old school DR stuff is lost in the world of the cloud and startups. It all comes down to how much in dollars or reputation the loss of a service costs the company...if you can quantify that and the number exceeds the cost of mitigation, businesses would be stupid not to put something in place to mitigate it.