"data is lost even with some redundancy" tl;dr periodic hash challenge verification of file integrity also serves as a trigger for file recovery from other nodes. all files are at least 3x redundant on the network, but you could set your file redundancy to 500 if you wanted to protect it from Armageddon.
but Storj doesn't know what is real redundancy and what isn't. You might think the file is backed up in 500 places but all those places end up being the same guy and maybe even same location. Locations can be disguised right? "sybil attack"