I know where it comes from, and 20 years ago it may well have been appropriate. The "use different" approach introduces a strong bias toward dependability at a definite cost in security while misoperation shows that we, as an industry, really need to focusing on security. We see that the vast majority of protection system misoperations are unnecessary trips. And of those unnecessary trips, the single most common cause is setting errors.
If I'm setting two different relays I have twice as many settings to work with as I do when I'm setting two of the same relays. If my greatest risk is a bad setting that causes an unnecessary trip I've just doubled my total risk of a misoperation. When the numeric relays were new and unproved, it may have been very prudent to take an approach of "if this one doesn't trip, hopefully that one does". But that's not the problem the industry has today. I've heard told that at one point, particularly when event analysis was far more difficult, that "well, it reclosed, so that's good" was a common attitude. Now we have PRC-004 and the requirement to analyze every operation to ferret out all of the misoperations and correct them. Security failures seem to be very low hanging fruit and, to me, the lowest of that low hanging fruit is the security risk associated with using different relays.