the motherboard capacitors start to swell or fail. I will go with five years.
While electrolytic capacitor failure is inevitable in the long term, you can't put a reliable sunset date on it unless you're talking about a specific piece of equipment with known failure rate data.
First off, we can set aside the old "bad electrolyte" stuff. The industry has known about that problem for decades now, so you should no longer run across that, except possibly from vendors squeezing every last penny from the BoM. MikroTik are optimizers, but my sense of their design esthetic doesn't have them pushing things that far. If it were otherwise, we wouldn't be seeing the occasional threads like this one, where someone brings up still-running 10+ year old devices.
What matters in practice are the capacitor design ratings and the thermal environment they're used in.
The two key ratings are the maximum temperature rating and the expected lifetime at that rating. A really good electrolytic will be rated for 5000 hours at 105℃. This means if you run it right at the thermal limit, it's expected to hold within its specs for 5000 hours, or about three-fifths of a year. It doesn't go poof-bang not-a-capacitor at the 5001st hour. It's simply allowed to start drifting out of its specs past that point. To the extent that the design has margin, you can keep running it after that point.
If there are EEs insane enough to design a device to run right at their capacitors' thermal limits, it's still likely to have some design margin, so a device based on such capacitors might still run properly for years before failing.
This is why wall warts fail so often: cheap capacitors packed tight up against power semiconductors in a small unventilated box.
You may then ask, "Why do most electronics last well over a year if the best caps are rated for only 5000 hours?" It's because most designs
aren't run right at their capacitors' thermal limits. The rule of thumb is that capacitor life doubles per 10℃ drop.
Let's posit that MikroTik put a crappy 1000 hour @ 85℃ cap in those RB1200s you have there, but the thermal design is such that even though the CPU runs right up near that temp, they put the caps far enough away from the CPU that the ambient air around the capacitor is more like 45℃. That's four doublings, 2⁴ = 16, so we can expect about 2 years of run time before it starts to eat into whatever design margin they've baked in.
To bring this thread back home, you can't say the RB1200s are due for replacement without failure rate data. I expect you're unlikely to get that out of MikroTik, but you have a fair sample size there. Have any of the RB1200s your customer owns failed yet? If not, why would you assume they're doomed? If so, what's the failure rate looking like?