There is a possibility that on a large multi-PM system, the cpimport.bin job, when finished, closes the socket to the splitter while the splitter is still trying to read. Or it may just send a 0 length message. Either will raise a CONN_FAILURE critical alarm and leave it up forever.

Nothing is logged in this case, which also needs to be corrected. If the alarm is raised due to a socket closing, an exception is raised, but no log is made, just the alarm. Exceptions should be logged.

So this bug consists of two components:
1. Anything that raises an alarm should log as much informations as possible. (For this bug, just the CONN_FAILURE alarm)
2. Fix the communication between splitter and PM such that a socket isn't closed prematurely.

This is a difficult bug to reproduce. Since it has no operational impact, it is marked as minor.