The Shareplex Admin Guide for 5.1 and up has this info in them as below:
"If you have SharePlex configured to start from a startup script and it returns an error similar to "System call error: sp_cnc (connecting from cdrpx.mcit.med.umich.edu) Bad file number Can't dup2 stderr," try adding nohup to your script. The nohup command directs a command (in this case the startup of SharePlex) to continue in the background after the current user (the script) logs out. The syntax is:
# cd /product_directory/bin
# nohup ./sp_cop &"
Though the Admin Guide does not delve on the reasons for the occurrence of the error, the most likely cause is that for some reason Shareplex is not able to direct the stadard error at Unix level due to some configuration problems with OS. The problem can also be memory related. When you run with nohup, it makes the process run in the background even though you log out. When you log in again, maybe the share memory is not able to reallocate for whatever reason.
The nohup should be able to get around the issue and is more of a permanent cure. Another thing that would help is to kill sp_cop and its child processes and relaunch sp_cop. But the error can recur. If using nohup, the chances of recurrence are extremely slim.