Why does DNS cache expiration need to be in the critical path?

Instead of a browser doing

1. Local DNS lookup (resulting in expired entry)

2. DNS query

3. DNS response

4. HTTP request

why not do

1. Local DNS lookup (resulting in expired entry)

2.1. DNS query

2.2. HTTP request

3. DNS response

4. If DNS response changed and HTTP request failed, HTTP request again

Maybe use two expiration lengths, one that results in flow 2 and a much longer one that results in flow 1.