Ok, here's an update after another attempt.
I edited the model (which is not actually flawless, having lots of little holes etc.) in Tinkercad, to make the support beams a bit shorter. Then I changed to 0.2 mm layers and z-hop of 0.2 mm. Following pm_dude's advice, I put combing back on.
Good news is, this printed with only a single support breaking off. Following the print inbetween cooking dinner, I even applied a little superglue to put that broken part back on, resulting in a molecule print that is at least usable for its chemistry teaching purpose. It will have to be painted in color, after all, so tiny errors are acceptable.
However, I also had a surprise:
- I cut out a detail of the model that looked particularly prone to failure, containing just four beam supports. The result was a really good print, with straight pillars and great detail. But,
- printing the entire model, which has 40 support beams, many layers are displaced horizontally, giving really ragged edges and a much lower quality overall.
It seems that my printer can't cope properly with precise repositioning when the extruder needs to move across a long distance. Surprisingly, the columns at the outer extremes of the print -- far left, far right, back, front -- are smooth and precise. The ones in the middle are ragged and rough.
I'd of course appreciate any hints as to why this is happening
(New images of the latest print + the detail cut-out with good quality, added to the album:)