This pattern:
t33: v8i32 = insert_subvector undef:v8i32, t35, Constant:i64<0> t21: v16i32 = insert_subvector undef:v16i32, t33, Constant:i64<0>
...shows up in PR33758:
https://bugs.llvm.org/show_bug.cgi?id=33758
...although it doesn't make any difference to the final result.
In the affected tests here, it looks like it just makes RA wiggle. But we might as well squash this to prevent it interfering with other pattern-matching. I think there's a similar opportunity for extract-of-extract.