I need to double check the implementation, but I think that someone waiting on the lock object in another process already has an open file handle to the file being deleted