My admission for the day: I have never relied on HashBytes (with the key as a salt) - I always end up get nervous and comparing all the values instead 😀

Technically I think the chances of a collision using HashBytes (with SHA1) in the way I describe is overwhelmingly unlikely, though the mathematics to demonstrate that are far beyond me. My co-workers consider CHECKSUM with a PK salt to be sufficient (please don't get me started on that).

Summary: when hash indexing, I use CHECKSUM to narrow the search followed by an explicit comparison.

To detect changes, I compare all column values (unless I am fortunate enough to be working in 2K8 when I try to use CDC). Many people swear by HashBytes - I am more likely to swear at it - though my views are not universally shared, it must be said.

Paul