That's up to you. My thinking about this though is that if a file is downloaded 'corrupted' then it is different from the files that you downloaded from. Therefore the hash of your file is now different from the original. This matters when someone else tries to download this file and finds two different versions of the 'same' file and has to choose one or the other.

You can't do a swarmed download for files that are different, and a corrupted file is different. If you carry this out on a large scale then you have several versions of the same file that have to be downloaded individually and at a slower rate since you can't swarm download the different file versions.