Handling non-homogeneous core systems?

On a multi-core system, OpenSBI currently cold-boots on a first core chosen by the "cold-boot lottery" (i.e., the cold-booting core is chosen at random from a pool of eligible cores). Then, once the cold-booted core boots, the remaining cores warm-boot with the assumption that their own set of implemented RISC-V extensions and other properties are exactly the same as the cold-booted core. In other words, OpenSBI currently assumes that the system's CPU is homogeneous.

I've been contributing to the Keystone enclave project, and we've been experimenting with taking advantage of systems that don't have a homogeneous set of cores. We've been having trouble, as expected, with booting into a system with such a property -- do you guys have any thoughts on how we might be able to handle a non-homogeneous system with OpenSBI as it is now? Or would we need to wait for a change to the boot process of OpenSBI?

No source files or tests are identified in the issue. Start by reading OpenSBI's cold-boot lottery and warm-boot assumptions described here, then determine how heterogeneous RISC-V extensions and core properties should be represented. Done would require an agreed boot-process design and corresponding implementation scope.