Do not cache webapp IP in avatar servers

There has been avatar outage since 15th of May when I killed last of the old webapp instances.

Error access logs

The apps nginx caches the reference to webapp servers and doesn't refresh that reference at all. The problem and the fix were the same as in https://gitlab.com/gitlab-com/gl-infra/infrastructure/-/issues/6971

This issue is to stay open and track work on the apps servers to be able to handle the scaling of webapp servers