Comment by trynotsober

When replaying customer queries against the next version, how do you compare results for queries using now() or approximate aggregates? Curious how you separate expected differences from actual regressions.