At startup time, each changelog server establish a connection with each other
changelog server.

At regular time each changelog server checks if it knows some changelog
servers with whom it is not connected and if necessary tries to open a connection
to this server.

This check is based on the hostname of the server.
When a changelog server uses DHCP this hostname is not always the same on
all the servers and this can result in tentative to open duplicate connections
which then result in spurious errors message.

In the long term, the changelog servers will use a discovery process and
this will be a good opportunity to discover the server ID of each other changelog
server and this will be a much better way to check for missing connections.

In the short term I've modified the code to use the IP address instead of the
hostname as an identifier for the remote changelog servers.

This should work fine with DHCP.