Originally Posted by thompr
It might even have to do with the fact that these days were on the weekend as opposed to weekdays, which follow different rules for recurring alarms. I wouldn't be surprised if there weren't some obscure bug in the "do alarm / don't alarm" logic involved with recurring/non-recurring, weekday/non-weekday, pathways which only gets tripped under very peculiar circumstances.
Some bugs, like Y2K, are not so much programming bugs as they are paradigm problems that result in fairly "sensible" bugs. By that, I mean you can make sense out of them by just observing the symptoms and understanding the principles involved. But some other bugs cannot be understood in this manner. Some bugs are just plain screw-ups by the programmer that result in strange manifestations. When you face one such bug, you can't just stand outside of the box and work your way to the explanation. You have to dive into the code. That may very well be the case here.
This may very well be a Y2K-like bug. Perhaps the program is assuming that the alarm set is for the current year (when set) so if you set it on Dec 31, 2010 for Jan. 1, 2011, it was really getting set for Jan. 1, 2010. That would be consistent why if you set the alarm now (like some people are saying) then it works. This would also be why they are working on it still...to fix the problem for any other alarms still to come and this might be a problem next year if you set your alarm in 2011 for a date in 2012.
I know this is just a guess, but it seems consistent with the symptoms...