Using robocopy with out the /MIR switch.  Approx 6TB, manually recreated shares.  1000s of files are being skipped, assume it it because users are online.  Prior to go live, plan was to close all sessions and files and do one last robocopy. 

What is the easiest way to prevent people from connections be re-established.

robocopy switches are

robocopy "<source>" "<destination>" /E /SEC /COPYALL /V /NP /NFL /NDL /MT:32 /Z /TEE /R:30 /W:10 /LOG:E\log.txt

Usually migrations like this are done in to steps:

1. Do the initial copy. At this point, we do not care about files skipped/failed. The goal is to copy most of the data, not all.
2. Do the differential copy. This step is done during a maintenance window. You stop Server service on the server and run RoboCopy. No users can access any of the shares on the server, so all the files are copied. At the end of the maintenance window you would probably would like to update logon scripts/mappings/DNS records/etc. so all the users would be pointed to the new server.