The two cliched sources of this problem are 1) someone pushed something out over the holidays that could have waited until January, or 2) peak capacity was negatively affected since the last time a spike happened, nobody had a way to monitor it, and so this has been broken since the end of May. On further reflection, someone will admit that they noticed a notch-up in response times and did not connect the dots.
