This appears as a weak signal across multiple tests and testbeds. The best signal is from soak test [0]. Bisect attempts were not very convincing, but the three perf-verify runs up to [1] are reliable enough for me, all pointing to 40745 [2] as the cause. Reading the commit message, I believe that was an intentional trade-off to improve performance in many-VRF configurations, so I think small performance drop in single-VRF scale test is acceptable.

[0] https://csit.fd.io/trending/#eNp1UEsOgjAQPU3dmEnaGoSNC5V7mFoGafjVtqJ4eguJGViw6Gf6fp3xoXd489icWHJh6YXJ1BRxY4fzPh6DtSA7MPoDgvMHSiswE1w_QQ0lGHuENLuD0IChmqq4vFYNSs5rcF0Bvlf1ZCevk13xCitvQmw1ErKZSHzlUJEgxhIU0C9SNj5G9NKpFr35ImliT4TrOCCChF4HhdEu0H-3ac6SfNf1rp3HGu--6t8QnFGNn5-iei4mdpL_AF-Ya_I
[1] https://jenkins.fd.io/view/vpp/job/vpp-csit-verify-perf-master-ubuntu2204-x86_64-2n-icx/117/
[2] 40745: fib: improve ipv6 fib scaling | https://gerrit.fd.io/r/c/vpp/+/40745