I think the point is, it is suboptimal. You’d get better results if you minimised the contextual splits for natural moments of completion rather than unnatural

I think the point is, it is suboptimal. You’d get better results if you minimised the contextual splits for natural moments of completion rather than unnatural moments of interrupt. This holds true in computing theory, feeding 10 things to a single core in sequence is more efficient than feeding 10 things at once and having the CPU handle interrupts.