> Normalization is not necessary in an immutable data set, however. The only reason to normalize immutable data sets may be to reduce the storage necessary for them. On the other hand, denormalized data sets may be easier and faster to process as inputs to a computation.

I disagree here. To derive new data or from a immutable data source you still want normalized data. Denormalization can happen at some point for processing. Sometimes normalization is not feasible, because it often requires modelling/design. But it is the best general case scenario for data to be well-structured and normalized.