How Moving to Event-Driven Architecture Solved Peak Load but Strained User Trust

A construction check-in platform moved from synchronous processing to an event-driven model to handle morning traffic spikes.

Using Azure Service Bus and CQRS cut peak API latency from seconds to under 40 milliseconds and prevented database overload.

But eventual consistency made supervisor dashboards lag by 5 to 10 seconds, causing workers to think check-ins failed and rescan repeatedly.

The trust issue was later reduced by showing an optimistic pending state in the mobile app UI.