Thanks for that analysis of our conflicting data about Mongolia. On
balance I'm inclined to agree with you that part of our 2008 change that
was at the time marked "this is almost surely wrong" is best reverted,
and installed the attached patch which incorporates your analysis and
takes your first proposed solution, which is simpler. I had one quibble,
though:
 

That news article doesn't cast light on the matter, because when it was
written the polls would have been closed in the central and eastern
regions of Mongolia regardless of whether Mongolia had two time zones or
three. So I omitted that citation in the commentary of the attached
proposed patch, which I installed in the development repository.