Why do coroutine-heavy systems still suffer from latency spikes?

Latency spikes occur due to blocking calls, dispatcher starvation, GC pressure, and external dependency delays.

Even though coroutines are lightweight, they still depend on thread pools and external systems. Latency spikes happen when threads are blocked by IO, when too many coroutines compete for limited dispatcher threads, or when GC pauses interrupt execution. External dependencies (DB, APIs) often dominate latency, meaning coroutines only expose bottlenecks rather than eliminate them.