Many CPU/SoC systems could be made vastly simpler while maintaining binary compatibility if you're willing to sacrifice performance/area. Anything that's not visible from the ISA level can be implemented however one wants. You can cram more logic per pipeline stage, stall instead of having complicated bypass logic, etc. Much of the complexity of modern circuit design comes from the desire to eek out every last percent of performance from a process, which is completely unnecessary for a historical project like this.