Bump the station build to v1.0.4, reduce live WAV chunk size to 0.5 seconds, and keep native speaker queues fed during streaming instead of draining after every chunk so playback is less jittery and stays closer to real time.
Add a station build version counter, show it on the monitor and terminal header, log it at startup, and report it to the server during register and heartbeat so deployed client versions are visible during troubleshooting.
Increase live WAV chunk size to one second and log chunk acceptance, queue depth, and retry state on stations so the HQ speaker path can be diagnosed from runtime logs.
Slice realtime WAV audio into small self-contained chunks on the server and feed them into the HQ speaker WAV queue on stations so music can start playing before the full file finishes transferring.
Move live music streaming back onto the proven DFPWM playback path and silently ignore legacy train_arrival, train_departure, and play_sequence messages from older station clients so they no longer spam logs.
Delete the station-side arrival/departure announcement requests and drop the server handlers for train arrival, train departure, and play_sequence messages so preview audio no longer interrupts live music.
Move music playback off the HLS/download-first path and stream 48k mono PCM chunks over WebSocket so stations can start playing immediately. Add station-side PCM chunk playback support and stream lifecycle logging to make no-audio issues easier to trace.
Henry_Du