Some quality checks are permanently not-possible on this rig because the 4-bin FFT cannot supply the required feature

Three quality dimensions are permanently not-possible rather than merely blocked on L3: onset-based density (events per second needs onset detection, absent from the 4-bin FFT), kick-versus-bass collision (needs onset plus sidechain detection), and timbre/centroid/mood features (need spectral centroid or richer analysis than four bins provide). These are architectural limits of the audio bridge — adding a perception layer does not enable them; only upgrading the analyser to expose richer data would.

Examples

Onset density (events/sec) cannot be derived from a.fft[0..3]. Even after an L3 perception bridge lands, this check stays not-possible unless the analyser itself is upgraded to expose onset timestamps.

Assessment

Distinguish [L3]-blocked heuristics from not-possible heuristics. What would need to change architecturally for onset-based density checking to become possible?