By all means offer fall-backs to Google DNS if it's not behaving correctly, for the reason you mention.
I've found it's also, quite poorly implemented, particular on CC Audios, I had an instance last year where my internet connection went down at my ISP, my DNS saw 10s of thousands of DNS queries per-device from each of my Chromecast Audio devices in the time I was out at work. It was almost 40k DNS queries in ~12 hours, per device.
Almost everything else on my network behaved normally, but the Google devices just went mental spamming the network with insane numbers of impossible requests, back-offs are a thing, they should use them.
Like somebody said below, this is pretty much a non issue unless google starts to force DNSSEC and certificate pinning.
I think it's fine to want that configurable and maybe, maybe even expect or ask for it.. But I believe concocting conspiracy and assigning malicious intent so readily, as many are, displays either an ignorance of how products get built for the mass market or a feigning of such. As others who have worked on IoT devices have pointed out, there are rather valid user experience concerns behind baking these "preferences" into the products.
> As others who have worked on IoT devices have pointed out, there are rather valid user experience concerns behind baking these "preferences" into the products.
Which is something that I have not denied. What I assert is that there absolutely needs to be a way to change that behavior if users wish. The lack of that ability is, in my opinion, something that is harmful, both technically and socially.
There is no mandate or expectation that all DHCP clients always use the advertised DNS settings. If there was than alternate DNS services like 1.1.1.1 or 8.8.8.8 wouldn't even exist in the first place, as your ISP would be in complete control of your DNS settings unless you splurged for a static IP.
The cast client decided to use its own DNS settings as many clients do - you can override the DHCP settings on just about any general purpose OS, for example. Even though they are DHCP clients. Do you call Linux allowing you to specify a DNS server as "circumventing how the internet works at a pretty basic level", too?
There's a vast difference between "users can configure it to what they want" and "it's hardcoded and can't be changed".
Either that's circumventing how the internet works or it isn't.
Sounds like you agree that it's not, in fact, circumventing how the internet works?