I've started by solving your screenshot. My plan is based on the idea of using the CatalogStarGenerator (CSG) script to generate a synthetic template of just the brightest stars, synthesised with an exagerated psf width. This can then be used as a luminance template for your "bloated" stars. This can then be coloured using the chroma channels from the peaks of the target stars (this would be simplest with a script, but I'd planned to prototype it with some iterative PixelMath). The advantage of this approach is that the "bloated" stars will have a shaped psf, so will respond reasonably to histogram stretch operations. However, I've hit a problem: CSG limits the synthesised FWHM to 100arcsec; your image is about 128 as/pix, so your desired bloated stars will be bigger than this. While a higher resolution image would help the processing, it won't solve this limit. CSG is not one of the installed script sources, so I can't easily modify this limit. I'll keep thinking about it...