With a sufficiently robust RCT, I'd think that indicates an issue with the observation in the observational study -- i.e., there are some confounding factors in the real world -- rather than any sort of issue with whatever effect the RCT found.
Conversely, if we had a large observational study that was then contradicted by a robust RCT... well, I'd still be inclined to trust the RCT.
(At the same time, I think a good observational study is, by its nature, more applicable to one's own life. If an observational study suggests that $FOOBAR is good, then -- as long as there's also a solid RCT confirming that $FOOBAR is safe -- why not try out some $FOOBAR for yourself?)
Conversely, if we had a large observational study that was then contradicted by a robust RCT... well, I'd still be inclined to trust the RCT.
(At the same time, I think a good observational study is, by its nature, more applicable to one's own life. If an observational study suggests that $FOOBAR is good, then -- as long as there's also a solid RCT confirming that $FOOBAR is safe -- why not try out some $FOOBAR for yourself?)