Comment by wink

I can't put my finger on it but I think many people see this as two operating modes, for lack of a better term. The more monolithic and enterprisey your solution (and with bespoke DBMS), the more it leans towards anything more complex than an index and maybe a couple triggers - and the more microservices-y where every small service owns its own database (and only half of them are RDBMS) the less complex code is desired in the DB itself, because you're also migrating off of single instances/clusters a lot (just taking a relatively dumb data dump with you, or even just adding new replicas, ship-of-theseus-like).