The fragments given the nature of the shredder are all shredded slightly at different places and they are all mixed up - so you have to piece together like a jigsaw puzzle into a single copy (genome assembly).
However suppose there are lots of palindromic repeats in the document like the phrase "stressed" repeated 100 times such that they exceed your individual 100 lettered fragment, or is it the palindrome dopleganger "desserts" repeated 100 times? Or a mix of "stressed desserts" repeated in different ways 100 times? And also what of the phrases before and after "stressed/desserts," how do you place them since you are not sure about the order/orientation of this repeat?
This is the challenge of genome assembly. Now having more copies of the shredded document helps but some repeats are very nasty and aligning them in the assembly is not statistically tractable. And also they don't necessarily encode for relevant genes so they are often filtered out. Furthermore, they are arguably the regions of the DNA that is the most prone to change, since you can think of them as the "filler text" of the document; if an important part of the document is unintelligible, then this document wouldn't survive nature's QC process of sexual selection. However, unintelligible parts of the document don't encode important genes, so they evolve very quickly because they have a higher tolerance of typo's and mutations.
So in an population pool that people use to assemble and come up with an consensus genome (keep in mind the population pool will have different gene variation), so it may turn out at a particular base pair position in the document, there is so much variation that the genome assembler can't conclusively come up with an consensus sequence or a reasonable distribution of base pair (A, T, C, G).