java.net.SocketTimeoutException: Read timed out when nodes leave the cluster

I'm seeing a lot of SocketTimeout read errors when nodes leave the cluster.
The nodes return with in a few minutes of leaving.

The cluster has a primary and 1 replica.

We are using Elasticsearch 7.16.3 in Kubernetes and when nodes get reclaimed which seems to happen often then we see a bunch of the SocketTimeout errors.

The timeout is set to 2500ms which seems to be a very long time.

We are talking to the cluster using a restful call.

Thanks for any help.