I started an exploratory notebook to model the typical camera capture process using spectral data:
It is early days but I will continue poking at it. The first problem that occurred is that the IDT is something to apply, for obvious reasons, on RGB channels but we are currently rolling the white balance multipliers at this stage while depending on the demosaicing algorithm it is better to white balance on the CFA mosaic data directly.
3 Spectral Slices
D65 Rendered Image with Computed IDT