Self replicating robots - how realistic?

[Sky Captain]

That may not sound impressive, but it reduces large sections of the infrastructure you need to storage tanks and your ezy-grow microbial culture. Shit, you could imagine automated crawlers that chew up rock and digest it in a microbial loaded stomach vat before passing the liquor on to the electroplating intestine and it all starts sounding a lot like you've got a biological organism than a factory. Those metal leachers need oxygen, which means you'd need to keep them indoors on some inhospitable world, making atmosphere for them with something like a blue-green algae. Basically you'd need to get your own little ecology going, admittedly, that isn't easy.
If you're going to do it with anything, however, microbes are the easiest place to start.

Building the big stuff would be beyond their skill set, but if you could marry them with a 3D printer you could start churning out components.

Yeah, that sounds like something that could be based on technology although still beyond our current level, but at least understandable to us, maybe even possible to build at the end of this century, not your typical semi magical replicator nanobot swarm.
One could imagine a ship carrying several of such machines enter orbit around planet then scan the surface to determine the mineral composition and then deploys those machines in a place where there is the most favorable mineral composition.
There probably would be several mining robots that dig up various rocks then robotic trucks carry raw materials to refinery that with help of genetically engineered microbes leach out various metals then refine them and feed to 3D printer that can produce all kinds of componenets and then feed those componenets to assembly robots that build final product. Advanced AI would be required that runs the whole process and can solve all kinds of problems that would be encountered.
Still sounds like a fairly large system that probably would mass thousands of tons, but still that is better than carrying your whole industrial base with you together with an army of engineers and technicians.

[Spectre_nz]

Regarding atmospheric alteration using genetically engineered microbes; surely if one designs a microbe that can thrive in the initial environment, then one would be able to take advantage of the power of exponential growth? Sure, there may be other factors in terraforming a world that would slow down the process, and one might need to re-inoculate the atmosphere with a different strain to keep it balanced right, but considering the importance of marine algae in maintaining a breathable atmosphere, I shouldn't think establishing one would take too long if the required elements for it were mostly already on the planet.

Exponential growth is kinda expected with any self replication senario. Alas, all the interesting stuff happens at the same time as the exponential growth is destroying itself.
However, keeping microbes in an exponential phase is really quite difficult, unless its an environement they really love and it isn't changing too quickly. Unfortunatly, what we want is for them to quickly change the environment, and that's likley to knock them out of an exponential growth phase.

For a robust terraforming method you're more than likley going to need to import a whole ecology, as the interactions between organisms make for great positive feedback loops. Species A removes the waste of Species B, allowing species B to proliferate faster. Species A's population is limited by species B, but while it's dying as fast as it's growing, its also making a rich nutrient slime for species C, meanwhile, Species D is slowly breaking down rocks and Species E can survive because species C is buffering the pH for it and it can modify the atmosphere thanks to the minerals released by species D...

Yeah, that sounds like something that could be based on technology although still beyond our current level, but at least understandable to us, maybe even possible to build at the end of this century, not your typical semi magical replicator nanobot swarm.
One could imagine a ship carrying several of such machines enter orbit around planet then scan the surface to determine the mineral composition and then deploys those machines in a place where there is the most favorable mineral composition.
There probably would be several mining robots that dig up various rocks then robotic trucks carry raw materials to refinery that with help of genetically engineered microbes leach out various metals then refine them and feed to 3D printer that can produce all kinds of componenets and then feed those componenets to assembly robots that build final product. Advanced AI would be required that runs the whole process and can solve all kinds of problems that would be encountered.
Still sounds like a fairly large system that probably would mass thousands of tons, but still that is better than carrying your whole industrial base with you together with an army of engineers and technicians.

It would be large, but smaller than a steel mill. Keeping microbial cultures growing isn't fool proof, and requires more than just liquid in a bag, so, there are technical issues. But, your 'spare parts' can be 1kg canisters of freeze-dried bacteria you use to fire up a starter culture if the old one fails. Better than lugging a set of spares for an iron foundry...

Edit:

Actually come to think of it, to work at a comparable speed to a contemporary steel mill, it would have to be at least as large as a steel mill or larger. Once on the ground at least. Steel mills would be faster at turning iron ore into steel, they'd just be energy hogs while they did it...