init.d can create duplicate processes if pidfile lookup fails

[fooshards]

System info:

rhel6.10
init.d
telegraf-1.8.1-1.x86_64

Steps to reproduce:

Get telegraf into working state, exercise init.d scripts

root@testhost ~]# service telegraf restart
telegraf process was stopped [ OK ]
Starting the process telegraf [ OK ]
telegraf process was started [ OK ]
[root@testhost ~]# service telegraf start
telegraf process is running [ FAILED ]
[root@testhost ~]# ps -ef | grep teleg
telegraf  7765     1  0 10:42 ?        00:00:00 /usr/bin/telegraf -pidfile /var/run/telegraf/telegraf.pid -config /etc/telegraf
root      7821 29851  0 10:42 pts/1    00:00:00 grep teleg

Alter the pidfile to put it in a bad state (actual observed scenario was likely due to /var mount running out of space, getting the pid file unwritable, but in any case, the pidfile was out of sync)

[root@testhost ~]# vi /var/run/telegraf/telegraf.pid

Restart - note that the original process never got stopped, but a new process is started

[root@testhost ~]# service telegraf restart
Starting the process telegraf [ OK ]
telegraf process was started [ OK ]

At this time two pids are alive

[root@testhost ~]# ps -ef | grep teleg
telegraf  7765     1  0 10:42 ?        00:00:00 /usr/bin/telegraf -pidfile /var/run/telegraf/telegraf.pid -config /etc/telegraf
telegraf  7886     1  0 10:42 ?        00:00:00 /usr/bin/telegraf -pidfile /var/run/telegraf/telegraf.pid -config /etc/telegraf
root      7926 29851  0 10:42 pts/1    00:00:00 grep teleg

Note that 'start' actions are properly guarded

[root@testhost ~]# service telegraf start
telegraf process is running [ FAILED ]

[root@testhost ~]# ps -ef | grep telegraf
telegraf  7765     1  0 10:42 ?        00:00:00 /usr/bin/telegraf -pidfile /var/run/telegraf/telegraf.pid -config /etc/telegraf
telegraf  7886     1  0 10:42 ?        00:00:00 /usr/bin/telegraf -pidfile /var/run/telegraf/telegraf.pid -config /etc/telegraf
root      8065 29851  0 10:43 pts/1    00:00:00 grep telegraf

Expected behavior:

restart should guard against creating a new duplicate process OR kill the existing process

Actual behavior:

two identical processes are started, metrics begin to be duplicated, and/or as these processes pile up resources begin to become unavailable

Additional info:

This happened mostly because another agent was restarting telegraf a lot, and this issue was teased out downstream from it