Trust tends to be all or nothing in practice, which is why one error can flip a source from oracle to fraud in an afternoon. Neither position is accurate. Expertise is patchy and shaped by domain; the person who is excellent on one subject may be ordinary on the next, and being wrong once is what everyone who says anything eventually does. Discarding them entirely is just the same laziness pointed in the opposite direction.

When someone you rely on gets something wrong, work out the boundary rather than the verdict. Ask what they are actually good at, and keep reading them there.