Joseph Rushton Wakeling:
Forgive me if I'm missing something obvious, but as it stands I don't see how the R250/521 algorithm you pointed me to can be strongly pure.
Sorry, the R250/521 idea and the strongly pure idea are unrelated to each other.
but wouldn't that be a memory allocation nightmare?
For the strongly pure random generator we should choose a generator with a small internal state (let's say less than 5 CPU words, they get passed by immutable value).