We took this approach with the first version of a from-scratch hypervisor at my last job. Portions of the hypervisor were separated into different shared librar

We took this approach with the first version of a from-scratch hypervisor at my last job. Portions of the hypervisor were separated into different shared libraries and we wrote a custom ELF loader to do the relocations. It was interesting but ultimately it was scratched in favor of a large static binary.