How to achieve fault tolerance at the DNS level?

Question

Domain Name System

The scheme is as follows:
1) In the domain registrar's office, we configure cloudflare servers as DNS,
2) In the cloudflare DNS section, we configure 2 (or more) IP balancers (NGINX) that, by weights and response types (200,500), distribute the load among app servers, in in case of an error, do not send requests to the corresponding upstream server.
What I want to know:
1) Is such a scheme reliable, so that you remove it and add it?
2) How to achieve fault tolerance at the DNS (cloudflare) level without a paid subscription (cloud proxying (cloud) is turned off)?
Thanks
UPD: if you set up some kind of DNS before / instead of cloud, then this will also be a single point of failure. And if there are 2 DNS servers, then how, to determine which one of them is responsible at the client level? Or the client will determine this himself: if not available, then turn to another NS server, right? If so, then how can you understand at the DNS level whether the APP server/balancer is available, as well as at the NGINX level?
Maybe it's worth throwing out of the cloudflare scheme?
DNS -> cloudflare -> nginx -> apps
Should I use AWS services? How much more expensive will they cost for "iron balancing". That is, there are requests to be considered or traffic, or both?

Answer 1

Cloudflare, like any normal CDN, has its own DNS resolver. Why do you need something in front of him?