Just so we can monitor that the macro itself stays speedy and also that load time doesn't increase (especially for default_mode="disable" which people will use in packages)
coverage: 100.0%. remained the same
when pulling f5f2adb27cdf1eb7ea9e269933af5fff31c8c4d3 on asv
into 167412f19a7ca6dd4aeb82006572230e46b32021 on main.
Just so we can monitor that the macro itself stays speedy and also that load time doesn't increase (especially for
default_mode="disable"
which people will use in packages)