the problem is centralization. If your ISP's DNS goes down it only affects the ISP's customers
https://www.theregister.co.uk/2019/07/02/cloudflare_down/
Cloudflare has had a good chunk of outages over the years. Or at least their outages make news here more often than most any other CDN provider I can recall by a large margin, I haven't tried to get stats so my impression may be incorrect.
Having DNS ride on top of HTTPS makes things worse outage wise I'd expect as that is a pretty common data path. Cloudflare's CPU flare up last year I don't believe had any impact on their regular UDP/53 DNS hosting services (I was in talks with them at around that time to use them as a DNS provider).
I've run my own DNS both recursive(internal) as well as authoritative for my domains(external) since about 1998.
I'd be more open to DNS over HTTPS if there was actually a number of resolvers people could run on their own equipment. Last I checked I haven't seen any (one recent forum thread on the topic here someone pointed me to a product but it ended up being a simple proxy to an already existing DoH provider, not capable of serving DoH from say a local BIND installation).
Perhaps that situation has changed in recent months I am not sure.
I do fear the number of end user issues encountered as a result of split DNS on vpn systems where resolving some host externally results in a different address than internally and that behavior being intentional. I did something like this to block access to vulnerable CMS systems several years ago, if you tried to acess them externally they hit an address where the load balancer inspected the rule and only allowed very specific url patterns through, if you wanted to manage the CMS you had to be on VPN. If you tried to manage from outside you got a big warning page saying you needed to be on VPN.
Another similar situation recently where I adjusted internal DNS during an extended outage so that users on VPN could connect to the application, and users on the internet would get a maintenance page. An alternative solution would of been use host file entries for internal users but that is even more complicated. In fact I had to help one user deal with their host file entries from another similar event 2+ years ago that they never removed which was interfering with the new host file entries they were trying to use(tried to resort to host file entries for that user after all attempts to use DNS failed, only later did they disclose they had other host file entries already in place that were obsolete, so I just had them remove them all).
I'd wager in both of these cases Firefox's (and probably soon Chrome's and others) behavior will cause problems. The article mentions being able to push a policy, well good luck with that outside of very tightly controlled orgs(i've never worked at such an org in my 22 year career). Just another annoying thing to have to keep in mind when a user has a DNS related issue.
Already complex enough to try to get users to clear their DNS cache and/or browser dns cache/restart browser to get around DNS caching problems this will just make it much worse.
oh well, pales in comparison to the headache that will be the new SSL expiry issues, ugh.