What science can tell us about C and C++'s security

This has nothing whatsoever to do with science. You made no hypothesis rather based the “postulate” off the lower bound of your observation over a very small sample set.

Imagine this, write a C program where you write x million functions where each simply returns a single number (ie three_million = return 3000000) etc. then the codebase contains as many lines as you can imagine and up to int overflow (or use bignumber lib) and 0% of your bugs are related to memory safeness.

this is bad science