The problem you have over the medium term is that CNE is incredibly cost-effective, so much so that you need something like multiple-order-of-magnitude cost increases to materially change how often it's applied. The alternative to CNE is human intelligence; it competes with literal truck rolls. You can make exploits cost 10x as much and you're not even scraping the costs just in employee benefits for an alternate intelligence program.

What that means is, unless you can foreclose on exploitation altogether, it's unlikely that you're going to disrupt the CNE supply chain for high-caliber state-level threat actors. Today, SOTA CNE stacks are probably available to the top IC/security agencies† of all of the top 100 GNP countries. It probably makes sense to think about countermeasures in terms of changing that to, like, the top 75 or 50 or something.

I think we tend to overestimate how expensive it is for adversarial vendors to keep up with countermeasures. It's difficult at first, but everything is difficult at first; I vividly remember 20-30 extraordinarily smart people struggling back in 1995 to get a proof-of-concept x86 stack overflow working, and when I first saw a sneak preview of ROP exploitation I didn't really even believe it was plausible. As a general rule of thumb I think that by the time you've heard about an exploitation technique, it's broadly integrated into the toolchains of most CNE vendors.

Further, remember that the exploit development techniques and people you've heard about are just the tip of the iceberg; you're mostly just hearing about work done by people who speak fluent English.

† Reminder that customers for CNE vendors usually include many different agencies, invoiced separately, in the same governments.