From what I've found, the problem appears to be that 1/8 watt resistors were used instead of 1/4 watt. I've been working with Ray to get that changed.
Chris' calculations are correct. At 66 RGB lights, we're at about 4 amps. The fact that we can have 100 lights on a string and still be able to measure the current at less than 4 amps, indicates that there's a voltage drop along the line and bulbs at the beginning will get a full 12V and those at the end will get less. RJ mentioned that "The last led's on the string are dimmer than the rest." in the faq. However, I can't tell the difference. Those that can should stick with shorter strings. I'm sticking with 100. This also explains why its the reds mostly at the beginning of the strings that are burning out. They're the ones getting a full 12v and overloading the resistors.
Why the MR16s don't have a problem and the DSC do, I'm going to guess it has to do with the output from the two controllers. The DSC is probably outputting a full 12v under load, and the MR16 is probably putting out less, under powering the strings and therefore not burning out the reds.
RJ - we're using the strings from Ray listed in the wiki too. If you've had yours running that long, can you see if the resistors in your dumb strings look smaller or the same size as the ones in the smart strings? If I hold mine side by side, you can easily tell they're smaller without having to cut them open. I wonder if there were a couple versions of the strings made.