1. Proposed title of this feature request

Change CoreDNS so it doesn't send duplicate requests upstream

2. What is the nature and description of the request?

Amex is using a third-party istio that hammers DNS immediately upon ttl expiration and want us to work around it.

We see many (10-20) duplicate name resolution requests hit CoreDNS within a few ms.  They miss the cache, so CoreDNS forwards all of them upstream.  Multiply this out by all nodes in the cluster each running many pods with the same behavior, and you have a massive number of requests.

3. Why does the customer need this? (List the business requirements here)

Amex applications require short-TTL names for fail-over reasons, and the load from frequent requests that happen in a couple of milliseconds miss the cache and cause duplicate requests to be passed to the upstream resolver, which can overload it.

4. List any affected packages or components.

CoreDNS