That idea seems to be considered something of a moot point nowadays…
It has been asserted elsewhere recently in this forum that
a) it misinterprets (and for at least a long time has) the ability of the compiler to optimise large numbers of instead rules into a number of separate rulebooks that don’t all run for every action
b) it’s not likely to noticeably impact performance anyway on any recent system
Whether anyone has ever actually tested the validity of these theoretical claims in real code, I’m less certain…