Indeed, just recently I was locked in a disagreement with some developers over an optimisation directive for a database query they were working on. This directive, a part of the SQL language implementation, circumvents the DBMS’s locking mechanisms. All too often used by ill-informed developers as a notional “performance improvement”, forgetting that as volumes increase, the side effect is that queries get to see uncommitted data.

It’s fine if used very carefully, in very controlled situations, but this was applied by default. It’s unlikely too to be a problem immediately, perhaps, but in six months with twice the user base and three times the volume? Thankfully the long-term and considered view won through in the end, and they’ve ditched the directive.