I ran into the same problem so I created my own batch file on the Windows master to monitor the files on the unix server, we mapped a drive on the master. I haven't run into a problem yet doing it this way.
What are you master and agent versions? Did you upgrade the agents before you upgraded the master and did the file dependency work with the old master and new agent? Are the master and agent in the same time zones? How long are you requiring the file to be stable? Is the agent time in sync with the master (check the "Time Diff" column in the Connections)? I'm not sure if Tidal uses the ctime or atime to determine file stability. I'd check them all to see if any are not what you expect.
If it was working with a 5.0 master, but not now there may be some "residue" on the agent. Set the agent job limit to zero and wait for all processes to complete. Stop the agent process. Confirm there are no rogue processes still running. Delete the directory corresponding to the agent name in the /opt/TIDAL/Agent directory. Start the agent (the directory will be rebuilt). Reset the agent job limit.