When a restart on a resource in a resource group is requested, why does pacemaker restart the wrong resource?

Issue

• I have created a group of resources:

# pcs resource
 Resource Group: group
     database   (ocf::heartbeat:pgsql):     Started node1.example.com
     appserver  (ocf::heartbeat:tomcat):    Started node1.example.com
     webserver  (ocf::heartbeat:apache):    Started node1.example.com
     mailserver (ocf::heartbeat:postfix):   Started node1.example.com

• When I request one of those services to be restarted, pacemaker restarts a different resource instead:

# date && pcs resource restart webserver
Mon 13 Jun 12:12:47 BST 2016
webserver successfully restarted

# journalctl -f
Jun 13 12:12:49 node1.example.com crmd[19581]:   notice: Result of stop operation for mailserver on node1.example.com: ok | call=140 key=mailserver_stop_0 confirmed=true rc=0 cib-update=200
Jun 13 12:12:52 node1.example.com crmd[19581]:   notice: Result of start operation for mailserver on node1.example.com: ok | call=141 key=mailserver_start_0 confirmed=true rc=0 cib-update=201

• Why is a resource other than the requested resource being restarted?