How do you deal with eventual consistency in distributed transactions?

hey everyone, i'm wrestling with a common challenge in distributed systems: managing transactions that span multiple services. while two-phase commit (2PC) is an option, it often brings performance and availability issues. i've been exploring patterns like Saga, but implementing and monitoring them correctly can be tricky. what are your go-to strategies for ensuring data consistency across services when you can't rely on strong ACID guarantees? any battle-tested best practices or common pitfalls to watch out for when adopting eventual consistency for critical workflows