According to Fiji Government (http://www.fiji.gov.fj/index.php?option=com_content&view=article&id=2608:daylight-savings&catid=71:press-releases&Itemid=155) and Olson newest tzdata (ftp://elsie.nci.nih.gov/pub/tzdata2010o.tar.gz) and various other sources (http://firstname.lastname@example.org/msg07566.html) Fiji is going to end DST earlier than expected. The end time is 2011-03-06 03:00:00.
My windows 7 and Mac OS X has timezonedata as it was autumn 2010 and the end time is 2011-03-27 03:00:00.
If we test this date, then detection will fail in many systems. Although I believe that in areas where this information is essential (Fiji itself), updates are provided as soon as possible.
What this mean? We should take the transition dates from latest or nearly latest transition, because future times may (will) change. There are many examples that dst start or end is decided yearly basis and only few days before the change occurs.
The optimum may be past year, so now in 2011 the most ideal year would be 2010. Transitions has already happened and they are reported in tzdata and majority of computer users has updated their machines, so the detection is most reliable. Maybe we could take into account two or three past years, which could be fine way to be more specific and detect more zones than by limiting only one year.