When a unique secondary index has multiple column families, the extra columns in the key might need to be skipped. The method right now to skip those encoded keys only works on the types that the vectorized engine supports, leading to assertion failures.