We had this issue for a while now, but its always been very elusive and happened on overly managed workstations usually.
But just now, it disappeared on the linux pulse machine … we only noticed because pending job scanning stopped working, and hasnt been running for about 30 mins. After restarting pulse, we got repository / db errors. I noticed the deadline.ini is completely empty with only some default values.
Which version of Pulse is running? This was something that should be fixed in 7.1.0.23 and later.
OK we will update asap, it happened again overnight (7.1.0.17)
In the meantime, lock those perms down. 
Pulse needs write access
The latest builds sorted the issue though
It doesn’t really. We had this issue across all of our machines, so we eventually just locked the perms down everywhere and started enforcing the file contents using Puppet’s file template system, and no more problems since. Good to hear this should be fixed going forward though.
Aren’t these settings important (maintained by pulse on restart)?
PulseProcessID=5119
PulseListeningPort=35952
etc.
We did the content enforcing as well on the slaves / workstation, heh! Didn’t realize others had this issue so widespread as well.
I think starting with Deadline 7, the process ID and listening port are written to a separate pulse.ini file.