Because I want to write scripts with the minimal possible support overhead. That means striking a balance between the lowest common denominator, and productivity. Does (for example) ripgrep give me so much extra speed that it's worth committing to the overhead of installing it and keeping it patched up to date on thousands or tens of thousands of machines? With a dozen or more combinations of distro/version? What if the next guy has his own favourite grep replacement? What if we construct a suite of tools and all of them grep in a different way, and someone else a year or 5 years or 10 years later has to debug it? What if two or more of these tools decide that they are the one that gets to replace grep with a symlink to themselves? Having been there and done that it gets very messy, very quickly.
Note I'm not saying that grep can't be improved - but I am saying that it would be better to PR those improvements into grep itself rather than creating a new and incompatible thing. Or fork grep and ensure that the improved version is fully backwards compatible.