Funny, it's not that different to programming in Node or browsers. The JIT is awesome, but theres so many stories of fragile performance pits.

the concept is the same: do you do work in advance and bear the cost of storing it and maybe not using it, or do you do work as-needed and hope that the additional cost of trying to deliver immediately is less than the cost of storage and overrun? the answer seems to be the same, too, "depends on whether you have resources to spare and whether the environment is stable enough to count on immediate delivery"