I disagree with the entire premise that sending a whole second robot on a space mission is desirable. All that extra mass! And why humanoid? Not a reasonable form factor for a robotic space vehicle. Anyway, on a space mission, mass is everything. Why spend it on manipulators for fixing robots when you could build redundancy where you need it most? Your general purpose robot fixing robot still has to carry spares! Why do that when you can have built in spares in the first place, without the mass of a whole second robot?
Why I brought up space as it reminds me of the Apollo guidance computer story. The military had cutting-edge ICBM targeting technology, and within a few years, more powerful computers were in people's homes. Once a capability exists, it spreads. The idea that companies could successfully lock down self-repairing robots when the US government couldn't even keep ICBM guidance systems exclusive seems... optimistic.
And you're right about planned obsolescence in theory, but self-replication is fundamentally different from a smartphone. It's like having a genie that grants more wishes. Once someone demonstrates true self-replication even in a proprietary system then the cat's out of the bag. Open source communities, countries with different IP laws, hobbyists -- eventually someone will reverse engineer it. The economic incentive is just too massive. It's not like trying to copy a luxury handbag. It's copying the means of production itself.
As for space applications, yes, mass matters, but that's exactly why modular, self-repairing systems make sense. Instead of one giant redundant system, you'd have swarms of smaller units that can reconfigure, repair each other, and even combine into larger structures when needed. Much more mass-efficient than traditional redundancy.
And regarding humanoid form sure you're right it's not optimal for space, but it's optimal for working in human environments. Our entire infrastructure is built for human bodies. Any robot that can use our tools, navigate our spaces, and interface with our systems has an immediate advantage. They don't need to be perfect copies - maybe four arms instead of two, different joint configurations - but roughly humanoid makes sense for terrestrial applications. And space applications that are designed and built by humans or for humans.
The question isn't whether this will happen, but when. What's your estimate for when we'll see the first truly self-maintaining robotic systems, even if they're not fully self-replicating?
