I can tell you that I had some of worst (early) nodes from last year and even those I did get to work after extensive splicing and testing. Yes, a large number were bad.
I ordered more nodes and had far better luck, but still had some that didn't work right. Ray always gave more than I ordered, and in the end I ended up with a fair number of extras, but the time spent splicing was significant. I probably got a free string or so in the end.
I was in the end able to get everything to work, but the road was difficult at times --- and I definitely had many DIY moments where I wasn't at all sure I'd ever get it working.
Long story short, I literally felt many times like I was chasing my shadow, saw all the funky behaviors described above and then some ... and in many cases blamed the SSC, the nodes, the software ... just about everything ... and in the end it was usually user error or some slight oversight on my part., more often than it was the nodes. Yes, sometimes I'd find one node that was broken and removing it often solved all my problems in a hurry. Very often behavior up to that node was good and predictable, but behavior after that node was funky ... so when I started looked at where the problems started, it very often would lead you to the problem child node that needed to be removed.
I've also seen completely wild behavior when the wires at the end of the nodes or strip are allowed to touch each other ... and found that just cutting them off and dipping them in liquid tape worked worked on solving my flex strip nightmares.
I also got so good at dealing with problem strings that I would often cut the string in half, then test either end. Very often one side was fine. I would then cut the problem side in half and continue narrowing the problem down by a half string at a time.
I found that in some instances, a node would be capable of displaying colors, but was not capable of relaying them on to the next node, or would be "slow" or sticky. I also found some nodes would work fine inline, but did not work well as the first node. Again, strange stuff like that would drive you nuts, but in the end it usually boiled down to either user error or one or two problem nodes that just needed to be removed or put somewhere else in the signal chain. I honestly thing it works fine with 128 node strings, but the odds of a problem node are like doubled, so if you have one troublesome node ... then it makes the entire 128 look bad at first. Its almost like Ray needs a lightkeeper pro device that can detects these troublemaker nodes somehow.
Some nodes could do RED, GREEN but not BLUE. Other nodes could do say GREEN, BLUE but not RED. These nodes would often appear to work using the TH2010 tester, but in reality they usually were not. Sometimes the white would be "off color' because when one color wasn't working, the white was tinted and wasn't the same as the others. So eliminating those nodes was critical to success on other nodes in the string and sometimes didn't matter.
To me it was very helpful to create a set of 10 nodes which I knew were reliable and would leave those hooked up to a test board with phillips screw connections ... and then I hooked some extra 3-pin quick connectors on the other side ... and that made things easier because when things start going bad, you start blaming software, SSC firmware, nodes, you name it.
I would say that while the quality sounds likes its improved a lot, this is not a road for the weak of heart.