The SP1 audit highlights accounting for the max # of byte lookups per instruction as a way to defend against multiplicity overflow:
https://hackmd.io/wztOd455QKWf-K8LXh_Fqw#Defenses-on-Multiplicity-Overflow-in-mathbbF_p
We do have precompiles much larger than Sp1 (e.g. G2 ops), which may have more lookups per instructions, and invalidate this defense.
/cc @adr1anh