How do you handle API performance testing for internal services that have different load patterns?

Hey everyone, I'm really curious about how teams test their internal APIs for performance. Our services are pretty varied. Some get steady, predictable traffic, but others see these huge, infrequent bursts of activity, like at month-end or when big batch jobs run. So, how do you actually change your testing plans for those different traffic types? Any specific tools or ways of doing things that work well for internal APIs versus public ones? I'm especially interested in how you decide what 'success' looks like and pinpoint problems in a system with lots of connected microservices.