event sourcing vs traditional db, when does it make sense?

hey everyone, ive been thinking a lot about event sourcing lately, especially for services that need a solid audit trail or complex state reconstruction. it feels powerful for certain use cases, but the operational overhead and potential query complexity seem like big trade-offs. when have you found event sourcing to be a clear win over a more traditional relational or nosql database approach? are there specific patterns or architectural styles that lean towards event sourcing, or is it usually a judgement call on a case-by-case basis?