I've set up neidermeyer to closely match Lopp's annual node benchmark machine (since 2017) and am running identical scenarios using Core v32 and BS latest. I had to build Core without shani because it picks it up from the machine unconditionally otherwise, and the target doesn't have shani. That review work typically starts in November. I expect to have the two outstanding validation performance optimizations (view-based validation and parallel confirmation) complete well before that, but even without them we are significantly faster.
(edited)