download recipe caches error pages and failes until page is manually removed from cache
The recipe doesn't check the md5 until after it has downloaded
a file. If it find an error, it raises an exception without cleaning up.
In subsequent runs, it checks the cache, find something and checks what it
already has, erroring when the md5 doesn't match. At that point, the user is toast.
It shouldn't place the file in the cache until the md5 check has suceeded and extraction has succeeded.