No problems to record the microphone, connect the analyzer for a nice vu-meter, re-sample the massive amount of data to something we can handle (8Khz, Mono) using 'xaudio.js' from the speex.js lib and to wrap it into an appropriate WAV envelope.
Stopping the recorder seems to be a different story, because the recording process severely lags behind the onaudioprocess functionality. But even this is not a problem as I can calculate the missing samples and wait for them to arrive before I actually store the data.
But what now? How do I stop the audio-process from calling onaudioprocess? Disconnecting all nodes doesn't make a difference. How am I able to re-initialize all buffers to reate a clean and fresh jump-in point for the next recording? Should I destroy the AudioContext? How would I do that? Or is it enough to 'null' the createMediaStreamSource?
What needs to be done to truly set everything up for sequential independent recordings?
Any hint is appreciated.