Well, I would not say it is "crashing" 
I would expect stdin for a UNIX service to be connected to /dev/null. It is quite deliberate that when the stdin input reaches EOF then it causes logstash to exit (since it knows it will not get any further input). That will happen immediately for a service configured to use a stdin input. Many service managers will automatically restart a service that exits.
The combination means that the logstash service with that input will sit in a loop, exiting and restarting.