For fun, I ‘demosaiced’ without any interpolation by simply scaling the local average intensity of the RGB channels, then converting to grey scale. Kind of like a local white balance.
It’s a terrible way to demosaic most things, but because the feathers are essentially monochromatic, it works well on the problem areas. You can see artefacts in some other places.
I think this gives the most accurate representation of the actual image in the problem areas.
Certain areas of the feather detail pattern frequency and direction in this shot are generating ‘interference’ with the photosite filtration pattern on the sensor - MOIRE!
Moire is optical - and is therefore baked into the raw file.
If the heron moves even by a small amount, or the camera likewise, then the interference will either move somewhere else or vanish completely.
Fashion photographers suffer from moire A LOT and have a technique of slightly rotating the camera in their hands between shots or moving the camera backwards and forwards - any subtle movement of the camera will break the interference pattern.
The less megapixels on the sensor the less you suffer with moire.
An anti-alias filter in the sensor cover glass pack will slightly blur the image and so mitigate the risk of moire.
More megapixels and no AA filter equals more risk of moire.