Load test 'coordinated omission' flaw can dramatically misrepresent latency.

A common load testing flaw called coordinated omission can cause latency measurements to be deceptively optimistic. The problem occurs in closed-loop testing tools that wait for a response before sending the next request, which omits requests that would have queued during a service stall. This omission can cause reported p99 latency to be orders of magnitude lower than what real users would experience. The issue applies broadly to HTTP APIs, databases, and LLM endpoints. Experts recommend using open-loop load generators or applying corrections to get accurate measurements.