+ Job Log file becomes huge, so it's update slows down schedd itself ( I have to use it since condor_q fails to answer the queries all together )
+ Condor generates empty output and error files, and to the file system ( which is AFS in my case ) takes a lot of time to create them.