You seem to be misreading slightly.
It seems like it says the preterite expresses modal remoteness and the perfect makes further remoteness, and so they call it doubly remote construction.
Yes. In both of your examples, CGEL makes the point that the doubly remote construction is necessarily counterfactual ("If they had gone tomorrow" necessarily means "they're not going tomorrow", perhaps because they already went yesterday, and "I'd rather you had gone tomorrow" necessarily means "you're not going tomorrow"); whereas with the preterite alone, as in "If they went tomorrow" and "I'd rather you went tomorrow", they wouldn't necessarily be counterfactual. (However, CGEL also gives some present-time examples where the doubly remote construction is not necessarily counterfactual.)
It amazes me that perfect tense makes time distance not backward but forward for making psychological remoteness.
No. The remoteness here is not the remoteness of future time; after all, CGEL says this construction can occur "with present and future time" (emphasis mine), and explains in a footnote that "The double remote construction is not available for past time because the perfect construction is not recursive"; that is, the reason you can use the doubly remote construction with present and future time, but not with past time, is that with past time the perfect is already being used to express anteriority, so it can't be re-used to express remoteness (since *"if they had had gone yesterday" is ungrammatical).