Lambda the Ultimate

cf. this coverity thread, i'm wondering what your favourite thing / biggest hope is for PLT applied to programming-in-the-large. what are some quantitative things (random e.g. more better typing maybe e.g. dependent or effect)? what are some qualitative things (random e.g. formal methods)? perhaps in other words, what interview questions would you use to suss if somebody knew good p-i-t-l? i'd be curious to know what you think the limits of PLT are here; what about large systems is a non language issue (e.g. managing a large team of people)?