There have been two substantial but unsuccessful thrusts towards capability hardware in the last 40 years. I have in mind IBM’s FS (Future Systems) and Intel’s 432. They had in common the flaw that the designers forgot about performance. This bias was in context of a grand plan to move system complexity from software to hardware. It was an extension to the “semantics gap” which was pushing in the opposite direction to RISC. This is not a necessary property of capability hardware; the Plessey 250 was an even earlier performant capability system. There are rumors that the Plessey system runs still today in classified applications.
Both of these thrusts imposed restrictions on some efficient programming patterns. Restricting the patterns of viruses is good but these machines went further. One can argue that the 432 failed because there was not available for it a software layer ready to fit will on top of the hardware. Most programmers who contemplated such a layer were wary. The 432 never gained much software. IBM’s secrecy tactic may have thwarted their plans to develop such software internally. Many in IBM could see that the performance was inadequate.
Both systems attempted to role several revolutions into one, and both failed. I think they were both pointed in the wrong direction.