Yes, I confirm that fixed it for Julia 1.6.
However… I have around 30 similar loops in my package. Before I go hunt them all down, can I confirm that this is the “right” way to use @inbounds @simd. It seems like I see them scattered around in different places in different code bases.