Close response state before admitting standalone speech

When a speech service supports both model-generated responses and standalone scripted utterances, clear the model-response-active state at the model's end boundary. A stale active flag can suppress the scripted utterance's own completion event, leaving its text buffered until a later response. The transcript can then assign the later response's timestamp and place an answer before its question; sorting those timestamps cannot repair the underlying association. Reproduce the sequence through native speech serialization and transcript aggregation, checking that the scripted turn commits before another response starts. This applies to services that derive transcript completion from response state, and does not establish remote playback delivery.

Applies when: A scripted utterance after a model completion is transcribed only during a later response. Keywords: transcript ordering scripted speech response completion stale state audio aggregation