My point wasn’t to address whether any hashing technique might be collision-free, but that assessing that fact is not a good metric for the fitness of a given hashing technique.