The difference with a 3d moulder being that, instead of taking a couple hours to 3d print a mould, then stop your production line and manually install the new mould in place of the old, then start it back up again you could effectively instantly form 3d mould (via microactuators or whatnot), do a 15 minute production run and make a couple hundred parts, then move on to mass producing the next part you need with no break in-between. Your "factory" could be in full production mode nonstop yet have a single line produce many dozens of different types of parts over the course of a day.
When one thinks of space colony applications, it quickly becomes clear how essential such a thing will be. Even if you try to simplify, you're still going to have tends to hundreds of thousands of types of parts that will wear out with time. Let's say 100k unique types of of parts with a mean lifespan of 3 years - that's probably pretty realistic for a colony. That means you'll have to produce a new type of part every 15 minutes nonstop - with quantities varying from one-offs to the tens of thousands, depending on the part. Now think of how big your typical production line is and how much mass that means transporting from earth. Clearly, rapid production flexibility is critical! (same applies to all steps of the chain, including robotic assembly)
(and yes, I know a single moulder or whatnot cannot achieve all possible production jobs, real production lines involve many types of materials and many processes... it's just an example of a common production mechanism :) )
(as another side note, it should even be possible to make 3d moulds for metals. Carbon fiber cloth - or better, graphite fiber cloth - can tolerate temperatures hotter than many metals, but still has stretch and could be shaped with an array of actuators).