@simon are you ever worried at all that some team is going to optimize specifically for your pelican-bicycle test, in the same way that the ATI Rage Pro Turbo team optimized for the Winbench 98 benchmark back in the 1990s?
https://www.tomshardware.com/reviews/3d-winbench-98,57.html
gpt-6-rage-pro-turbo maybe?
