Yes, stuff like MTE will help, and what you mention is already being used on Android, see HWSAN, and there is now some movement regarding C++ security profiles, however there is still a very long way to go.
This is how Microsoft sees improving safety on C++ codebases,
"Build Reliable and Secure C++ programs"
https://devblogs.microsoft.com/cppblog/build-reliable-and-se...
"Lifetime Safety in C++: Past, Present and Future – CppCon 2023"
https://www.youtube.com/watch?v=PTdy65m_gRE&feature=youtu.be
This seems like an incredibly low amount of money to achieve something like that.
Similar to how clang slowed down as Apple and Google decided to focus on their languages, and key contributors abandoned clang.
That site still lists 2x CPU overhead. I am fuzzy on details but I believe some variant is in the works that has low enough overhead to enable in the wild. That site also just says it uses TBI so possibly the performant version I am thinking of is just a version that uses full MTE?
For hot parts of the code that the sanitizer can guarantee is safe, either due to it being trivial or with annotations(https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p27...), it can safely exclude these checks to limit performance impact.
This seems much more realistic than e.g. Microsoft rewriting Excel in Rust with bug for bug compatibility. It would also benefit Rust code, that has sections marked as unsafe.