There are five MySQL instances configured with a working group replication.

The cluster will shut down to preserve data consistency.

The cluster has built-in high availability and updates group_replication_ip_whitelist to remove the unreachable nodes.

The group replication will buffer the transactions on the online nodes until the unreachable nodes return online.

There could be both a 2 node and 3 node group replication still running, so shutting down group replication and diagnosing the issue is recommended.

A manual intervention to force group members to be only the working two instances is required.