I *do* like the suggestion Daniel Karrenberg made how to tackle this - give the NCC more liberty how to handle "experiments" by consulting, if needed, with an expert panel. I do see the issue in defining "expert", but maybe this could be made sufficiently lightweight - "ask for a volunteer group of individuals that have had hands-on experience with BGP routing for <n> years" (because, I think, that's really the crucial part here, to differenciate from other setups that can do the 50% just fine, or use RFC1918 space instead).
you are a (new) LIR applying for IP space. you submit an addressing plan. the ncc convenes a volunteer panel of your competitors to evaluate that plan. oops! tragically, research is competitive, and the ideas are the protein. [ fyi, i admit to being just a shill here. it was reg services who asked for help on the issue. ] randy