i think what is happening is you have the files on the node where the instance db01 is running, when you connect to user/pass@db it doesn't necessarily connect to db01 even if you run it from the node where the db01 instance is running. What was likely happening was the connection was made to db02 where the directory containing the files wasn;t visible (the files were not created on a cluster filesystem which is visible on all nodes). Even with cluster=n this can happen all that means then is that the job would still start on db02 say but would not attempt to run on any other node than that.
If you use ASM the file is visible on all nodes and you get no issue, also if you use a cluster file system it is visible on all nodes.
Hope that explains it?