That is a good point.

Also, I'm far more skeptical of the ability to correct for confounders than, say, Walter Willett is.  For example, once they found a confounder and adjusted for it and their results changed massively, and an interviewer asked them about this problem, and they just tried to play it down.  But clearly, getting the confounders correct is essential... and also impossible.

Nonetheless, I'm also not convinced that this is entirely explanatory.

Koreans are small.  It is well known that within a species, the smaller individuals live longer.  (Between species, the larger ones live longer).  So that one is confounded too. 

 well you can do controlled studies but those are expensive to run long-term or with large enough sample size and even in the best designed studies there is the issue of who is paying for them ...