DNS Propagation Does Not Exist: A Suggested Change in Terminology

A more concrete reason that the term "propagation" misleads people is that they're convinced that a DNS-based authentication token (like a TXT record that proves control of a domain) won't be visible to a validator until the record has "propagated", even if the validator has never checked that record before, or even if the validator doesn't use a cache.

I encounter this confusion all the time when helping people use Let's Encrypt, which can use TXT records this way.

There's another phenomenon which is a little bit more akin to "propagation" but which doesn't exist in all DNS infrastructure: especially if your DNS provider is very distributed or multihomed, there may be a synchronization time for all of its authoritative DNS server instances to find out about the new records. This is typically "push" rather than "pull", following the concepts of this article, but it's still not exactly what people are likely to think of as "DNS propagation": its speed is totally unrelated to TTL values, and it only exists for some DNS hosts, not all.