So I did some tuning on the fan shroud, moved the nozzles closer to the hot end and aimed the outputs towards the nozzle. I printed the axial version since getting the radial version to work might impose some limitations to printer's max Y size. Currently the fan sits under the X-beam, the axial version has plenty of room to be there. Radial on the other hand is a bit harder to get positioned properly. I looked at some fang-type shrouds in Thingiverse for inspiration, for example this: https://www.thingiverse.com/thing:2175956. But something like that will require some intensive modeling and reusing available models probably won't work.
IMG_20200721_102735.jpg

Anyhow, I printed couple of benchies and there seems to be a bit of improvement, but still has issues. First one is sliced with KISS, variable layer height 0.15-0.35, 0.6mm nozzle, fan 100%. I also reduced perimeter/solid overlap and retuned my extruder e-steps since it seems there's a bit of over-extrusion in some points, causing the inconsistencies of the extrusion.
IMG_20200721_102402.jpgIMG_20200721_102443.jpg

Next one is sliced with PrusaSlicer as I wanted to see if the slicer has an effect. 0.2mm layer height, 0.6mm nozzle, 100% fan. The default print speed of the profile I modified was bit faster than on KISS an that shows on the ships prow. Also PrusaSlicer has adaptive layers now it seems, I haven't tested that yet.
IMG_20200721_102427.jpgIMG_20200721_102451.jpg

I'll need to design a fan shroud that uses a radial fan, axial just seems to be too inefficient. Designing those ducts similar to the fang-design will be an interesting challenge with OpenSCAD, doable, but not sure if looking forward to it...

For this printer, I'd say the quality is good enough as it is since I mostly print mechanical parts that doen't need very high level of quality. Also with 0.6mm nozzle the layer heights usually are around 0.3mm or more so the small deviations probably just fade away. Still, would be nice to get this sorted out.