Could application-provided memory priorities complement Linux's existing memory management?

I'm not a kernel developer, but a hobbyist who has been thinking about memory management from a slightly different perspective. I'd be very interested to hear whether something like this has already been explored, and if not, what the fundamental obstacles would be. The idea came partly from working with Arduino-class systems, where a few...