Here is a failed attempt to build a scaling algorithm, but maybe relevant to this discussion.

The faces (originals in the upper right) were scaled to 3x, and I tried working out the math for the smoothing on half of each of the faces rather in order to have an easy comparison. The faces on the left have two additional colors inserted between each value step, the two large faces on the right only have one additional step added. And they look about the same to me.
This suggests if you scale up pixels by 3x, the color ramps don't need to have more than twice as many steps to appear smooth. Conversely, if you reduce the size of your pixel art by 3, like when you go from a 320 pixel screen to a 1024 pixel screen on the same physical display, you should roughly double the visual distance between colors in your ramps in order to achieve the same level of visual impact.
I say that with plenty of caveats. This is only based on this one gray scale image, so the ratios are probably wrong, and it may not apply to hue shifting. This was also done with RGB light values, so that's going to be a wonky scale.
I imagine that while modern monitors may have a greater range of real values (light output) than older models, the difference is not all that large. Which means the maximum effective visual distance between two pixels is a net decrease on modern higher resolution monitors. This supports the original assertion that pixel art looks better at 2x. Someone could do a school science project on this sort of thing, I imagine.
On the other hand, if retina displays (200+ dpi) become more common, a lot of this is moot.
Tourist