The real hitch is that most of the CRT “magic” is happening in your brain, not in the optics of the CRT. When you look at a CRT image, your brain is subconsciously doing a lot of smoothing, filtering, edge interpolation, etc. in a selective way based on your higher level understanding of what you’re looking at. For instance: If you see scanline gaps over solid blue water, your brain interprets them as waves. If you see scanline gaps over solid blue sky, your brain interprets them as foreground noise (like looking at the sky through chicken-wire-reinforced school bus windows) and filters them out.
It’s not possible to make a shader do that, or even anything close.
The other path towards the desired result is to simulate a CRT well enough to convince your brain to do all this. The price is that, while they do trigger this desirable psychovisual stuff, scanline gaps, and halation, and visible subpixels, and so on, all look objectively bad by every other measure. My personal preference is for just enough CRTishness to get the brain doing the psychovisual stuff, and no more.