Hi Mark,
Yes indeed, you can use SIMD to speed up BVH traversal, on the CPU that is, of course. In fact it is typically used to check multiple rays against a single AABB, so that rays that travel roughly in the same direction benefit from the faster calculation, as well as the shared cost of fetching AABB data from RAM.
I would love to write a bit more about that; perhaps I'll do a few ‘addendum' articles for the series. ?
Jacco.