dont use eventual consistency unless you really have to.

hey everyone, i've been wrestling with consistency models in distributed systems lately. we're building a new feature that deals with inventory management, and the trade-offs between strong consistency (ensuring all reads see the latest writes immediately) and eventual consistency (allowing for a period where reads might be stale) are really hitting home. strong consistency seems simpler to reason about, but it can really impact availability and latency, especially as our system scales. eventual consistency offers better performance and availability but makes the application logic trickier. what are your experiences? when have you found one model better than the other in practice, and what were the key factors driving that decision?