As time progresses, (AI) tools are going to be more powerful in detecting suspicious code -in any language-.
Once the Safe C++ spec gets done, one could make a lint (or whatever) version that supports checks to the specification. I expect this to happen in the near future anyway.
Abandoning C/C++ does not solve the problem today, or in 2026, or 2030. Some solutions architectures cannot 1:1 be applied given another environment. That is not just a rewrite - it is a fundamental redesign from the early lifecycle phases of a product, and it takes time, a lot of it, to catch up and overtake the mature level the product has today.
My bet is on better tooling supporting "memory safety" in the near future. It will cause a lot of headaches, but less than redesigning the product.