Is it possible to pull input from Audio Unit (v3) in excess of the requested frames to render in real-time, not offline? Or, is it possible to render some of the frames and wait for additional input? Either way would be fine.
By the way, the docs on Audio Units are a mess at this time, I find. A newcomer to the API is met with the challenge to filter through what applies to v2 and v3, and many doc articles are out-dated. To me, I'd say the docs of v2 and v3 should be separate programming guides — that would make it less confusing.