In 2010, I recommended ballistic-missile-defense systems run their servers under both Linux and FreeBSD to mitigate the risk of kernel-level remote DoS 0-days [1]. Now it looks like N-version programming has failed yet again [2].
But even granting that both affected OS flavors have serious DoS issues, a service that used a mix of OS flavors in its servers would be more resistant (not impervious) to attacks based on these TCP problems than a service using only one flavor. So do you really count this as a failure of N-version programming? (Honest question! This is not my area of expertise.)