It doesn't have to. It raises an error that the system can detect and take action on. Usually that'll be some combination of interrupt/reset and an external pin to let the rest of the system know what's happened.
What raises the error, and how does the system know that an error has happened? Like, if you have two processors calculating 2+2, and one comes out to 4 and the other to 5, how does the system know which one is correct? Actually, typing it out I think I get it now. It doesn't need to know which one is correct, it just has to redo the calculation if there's ever a disagreement. Then if somehow both processors calculate 2+2=5 simultaneously, the next computer over will disagree and everyone will repeat the calculation, and that's why they have 3 levels of paired redundancy and the chance of 8 simultaneous single-event upsets is low enough for their risk tolerance. Ok, now I get it.