Since you asked in a PM, here's how I do some of the CC stuff.
Most of it is pretty sloppy. I paint stuff like rockets, dropships, rocks, etc big. Then I scale down and sharpen a bit. Sometimes I fade the sharpen if I want less of it. I try different sizes for the scaledown (17.4%, 17.5%) and I also try and reposition multiple copies of the original at different places on the canvas. Why? Because of the sampling and filtering. Sometimes the scaler picks different pixels and it looks better or worse. I experimented once with writing my own scaler, which was inbetween nearest neighbour and bilinear, but that's another story.
Anyways, I want to make sure that stuff like... important lines etc. come out sharp. It's possible to use parts from different scaled down versions to piece the thing together.
Then I use a 1 px brush: Removing scaling artifacts and noise, flattening some areas, sharpening some lines which have become 2px blurs due to bilinear filtering. The edges needs treatment too. I use a different color to make sure I remove all aliasing on the outlines. Then I paintbucket it back to a neutral BG color to make sure I see the sprite in its natural environment. The underside of the sprite might need a darker outline, and the top, I might use the dark color of the shape + some blue to hint at a sky above.
The curves should be smooth and not have jaggies. I try to avoid staircase lines going 2,1,2,1,3,1,2,2,1 px etc. A curve should go something like ..3,2,2,1,1 or whatever. I also try to make sure I have nice dither patterns where I want some texture.
What else? I don't care about the palette much, because it's not NES sprites which only can have 3+1 colors. So, I use a 1px brush set on mid opacity to nudge rust, dents and highlights in (keeping some of the original colors). I also use it to flatten areas with high texture into lower texture areas, or the other way around.
Then of course I just lazily run the shit through my batch processing program which uses my own palettizing routine (just a 3D pythagora to find the closet match). This means that I don't care how mangled the sprite is after it comes out. The current CC palette is pretty bad, especially with the orange-yellow ramp and darks. This is notable on the Dreadnought which has a particularly bad banded look on the cupola. That's how it came out, and I didn't bother to hand optimize it. Why? Because I still have the 24bit original, and if I redo the palette I can just run that through my batch processing program and get a better look.
Also, since the rotation in CC is unfiltered, most pixels are mangled anyways. What matters is the readability of the big shapes, and the general design of things.
1:1. I keep things rough because tiny details won't survive the scaledown. They need to be handmade at pixel level later.
These are not the final ones. I think there will be some different text stuff. As for whether these are 'pixel art' or not, well, they're pretty borderline. There's just a little hand optimization, mainly around the edges, and at key points. The cog needed some work. They don't need to be super clean because they're meant to be viewed at 1x.
Edit: fixed image link
