> Seems like the consensus is that engineering interview is broken, but no one has come up with a better method?

These sort of "consensus"-es appear for most sorts of screening processes where the majority of people are screened out. See: any form of standardized testing.