I would also submit that Kelvin is not an appropriate reference for the colour of LED lights. Kelvin only makes sense when discussing blackbody emissions, it is essentially a description of temperature. LEDs do not create light via simple heating. Their spectrum is not the smooth blackbody curve that the kelvin scale suggests. It is in fact a combination of several strong emissions at narrow wavelengths that combine to fool our senses into perceiving white light.
Proper discussion of harmful emissions from LEDs should instead speak of intensities at a particular wavelength and perhaps ratios between wavelengths rather than an outdated reference designed for blackbodies.
While I agree with you from a standpoint of accuracy, Kelvin is still really useful. It also happens to map pretty nicely to how we perceive color.
The N-step MacAdam ellipse on a 2D CCx/CCy plane is what gets used when you dig into tint, but people love to reduce complex systems to a single number. And if you want a single number, temperature is the best we have.
I'm curious what your suggestion might look like in practice because it would be even more complex than the MacAdam ellipse.
In practice, I could see a simple description of the intensity of visible along with a percentage of the 'bad' light within that intensity. That is an analog for how our eyes actually function. We have a retina that sees specific wavelengths, but an iris that regulates the amount of light hitting the retina based on an average intensity across the visible spectrum.
Harm comes to our eyes when 'unnatural' non-blackbody light plays tricks with the system. Looking at an eclipse isn't harmful because of some new light that only exists during an eclipse. It's because the darkness allows our iris to open despite the presence of harmful UV light. (The same goes for vintage sunglasses that don't block UV.) So to discuss the biological safety of light one needs, at a minimum, both the average intensity to judge how open the iris will be, and some reference for how much harmful light is present. Imho that can be expressed as an intensity and a percentage. An eclipse is a low visible intensity with a high percentage of harmful light. These blue-heavy LEDs could be similarly described.
CRI accounts for this, though it does not reveal the specific wavelength components. You can however be confident a 90+ CRI light source is not going to have an unexpectedly high intensity of any given wavelength.
Proper discussion of harmful emissions from LEDs should instead speak of intensities at a particular wavelength and perhaps ratios between wavelengths rather than an outdated reference designed for blackbodies.