the satellites are life limited whether they have noticeable faults or not (mostly by the remaining propellant, but also by TID radiation and payload obsolescence), and any sensible repair-in-place mission aims to proactively repair satellites near the end of their practical life rather than reactively responding to issues. The robotics (and fluid transfer for refuelling) are basically there, you just need satellites with an architecture that enables repair/replacement/refuelling to move beyond demonstrators. And most satellites are not built in that way.
As for whether it's worth it, it depends a lot on the satellite and the orbit. Starlink: probably not worth upgrading. Geostationary satellites: there's a reason why what goes up there commands oldspace prices (long way up, intensive radiation environment) and that reason also justifies serviceability. Hypothetical LEO datacentre with most of the mass being solar panels and radiators that last a lot longer than the chips are commercially relevant: yeah, probably makes sense to have the chip payload swappable and the thrusters refuellable.
A mission for one satellite to repair a lot of others is complex (it's basically an extra-hard version of the travelling salesman problem) but when the alternative is inserting many entirely new satellites into n orbits, sending a tiny fraction of the mass to n orbits via a reusable servicer starts to make a lot of sense...
Exactly, the scale and scope of payloads is increasing as launch costs are decreasing.
If starship launches ever get anywhere near the 2m-10m claim.. Suddenly fixing or retrieving that 500m-2bn satellite. Or rotating the compute modules on that 150bn space datacenter constellation..
Honestly I could envision regulation requiring it at some point.
As for whether it's worth it, it depends a lot on the satellite and the orbit. Starlink: probably not worth upgrading. Geostationary satellites: there's a reason why what goes up there commands oldspace prices (long way up, intensive radiation environment) and that reason also justifies serviceability. Hypothetical LEO datacentre with most of the mass being solar panels and radiators that last a lot longer than the chips are commercially relevant: yeah, probably makes sense to have the chip payload swappable and the thrusters refuellable.
A mission for one satellite to repair a lot of others is complex (it's basically an extra-hard version of the travelling salesman problem) but when the alternative is inserting many entirely new satellites into n orbits, sending a tiny fraction of the mass to n orbits via a reusable servicer starts to make a lot of sense...
If starship launches ever get anywhere near the 2m-10m claim.. Suddenly fixing or retrieving that 500m-2bn satellite. Or rotating the compute modules on that 150bn space datacenter constellation..
Honestly I could envision regulation requiring it at some point.