I can't believe I never saw this was yours.. we have things we should talk about. 

There are a lot of other ways to hone out your passages. Just make sure you never cook some pla in this thing or you'll never get it out short of firing it. That is not an option in my design but it might be in yours actually..

I eventually settled on straight through passages with a matching cutter precisely fitted to the depth so I could hone the entire cylinder without worrying about destroying the shape.

Have you tried an abrasive like glow-in-the-dark or carbon fibre for the inside core and a non-abrasive, transparent material on the outer? I suspect that this hotend could be used to print traditionally-abrasive materials without abrasion.

Should help, but still need to melt the abrasive in a separate channel which erodes overtime. Not as critial short term. Needs trying.

I recently found out about "extrusion honing", which is where an abrasive paste is pushed via pressure throughout a part in need of polishing. The consistency looks similar to molten filament, so abrasive filament might actually improve the internal channel surfaces.

Honing might be possible as a last optimization step. Though there isn't much material left between each color channel.

The transparent version 3 looks very nice. 

1) Is the design being printed in metal, or are you testing it in resin first so that it can be optimised beforehand?

2) Hypothetically, could an off-centered inlet approach improve coating performance? See [https://youtu.be/3sp91_g48jA?si=_ranJF1kadeDl4fQ&t=333] to show what I mean. This log [https://hackaday.io/project/190831/log/222620] is the project log I've written to further explain my hypothesis.

Some 'swirl' might help, but adding more and more channels also adds complexity. Coating works well enough already, so I'd probably focus more on reducing the number of channels and their length and the overall size of the nozzle. PLA doesn't do well if it is idleing for too long inside the nozzle.

I was envisioning 2 inputs for the coater (like your second version) for the swirl, not the 6 seen in the linked video. Potentially 1 could be used if the coater void is shaped similar to a blower fan.

I also agree on finding ways to reduce the amount of molten material present in the hotend. There is also the unexplored strategy of displacing idle inputs with a "cleaning filament material" channel.

Is there a reason why it looks like there's only 2 voids around the primary channel path instead of 3 in the Blender screenshot? Is there an internal Y connection between 2 channels and coating for the other 2 channels?

Yes, Y connection for two 'core' colors and two voids for 2 'coating' colors.

With that additional channel, a greyscale 3D printer now seems possible (White, Black, Clear/Alpha), and potentially an R->G->B->R... layer full colour print with the Y splitter idea I mentioned a while ago. I'd imagine the latter would look similar to book-edge art.

The first version allowed for grayscale too. White core + black coating with varying thickness, gives you different shades of gray.  

Current version has 4 primaries instead of 2, so already a much bigger spectrum. You could add of 5th or 6th if you wanted to maximize the number of colors allowed during a print. But if you only want to make one specific filament color, 4 channels is all you need.

Oh that's even better! I assume those colours are WCMY, since technically speaking, a white filament coated in cyan + magenta + yellow would theoretically appear black?

Yes, WCMY or KCMY, depending on the tone you want to achieve. C+M+Y is only black in theory, in realtiy most pigments mix to a dark gray color, so you still want a pure black at times for contrast...and contrast is more important than a big range of hues..at least for humans.

Would that apply to this process, which does not mix pigments but instead derives colour from the thickness of materal? This [https://www.youtube.com/watch?v=5inTryUaQyg] is what I'd expect to see, though I could imagine it looking slightly less contrast-y if against a white wall instead of illuminated from a bright window.

Coating or physically mixing paint is both subtractive. Example in the video is a bit contrived, appears black because filmed against a light source, it's more like a dark gray under normal condictions. Otherwise inkjet printer wouldn't use a K catridge.

Is it possible for you to explain how the hotend was built?