Audit retained turn timing before scoring voice response latency

Before scoring historical voice-agent response speed, verify that the stored events contain caller-turn boundaries and the corresponding audible bot start. A latency observer in source code does not establish historical coverage, and a component time-to-first-byte breakdown cannot substitute for caller-to-audio latency. Rolling event buffers can also discard early turns, so report capped calls and missing observations explicitly before comparing models. Validate reconstructed timings against recordings on a reviewed sample; until then, describe the missing coverage rather than publishing a response-speed percentile.

Applies when: Retrospectively measuring conversational latency from bounded voice-pipeline event logs. Keywords: voice latency Pipecat turn timing audio observability rolling buffer missing coverage