- IBM i / AS/400 / System/38's Machine Interface was made in 1978 before we knew many of the things we know about language and machine design today
- The JVM was made specifically for Java, even though it has mutated into supporting other paradigms
- I wouldn't be surprised if the CLR were actually backed by research like this, but does somebody here know?
- ParrotVM was made for hosting a range of language paradigms but, happy to be proven wrong, I'm kind of assuming it was designed (1) from craftsmanship, gut feeling and artistry rather than theory, and (2) not so much toward implementation efficiency
- LLVM IL, maybe? Research-based compiler with strong theoretical foundations, a focus on efficiency and hosting multiple languages, but maybe not with an eye toward a durable and standardized representation?