Measure audio drain before blaming synthesis for transfer delay

When a voice transfer waits for speech completion, compare generated audio duration with wall-clock playback duration and synthesis first-byte time. Fast first audio does not mean the announcement has drained through the transport. Backlogged output delays the speech-completion event and therefore delays the transfer itself. Measure event-loop scheduling lag and outbound control-frame residence alongside provider dial and answer timestamps. Residence identifies waiting, not the processor consuming CPU; container CPU headroom also does not exclude contention on a shared single-thread event loop. Do not name a dominant blocking function without profiling evidence.

Applies when: Transfers or other control actions wait for an announcement to finish. Keywords: voice transfer delay audio playback drain event loop speech completion