Hi Paul, thanks for your comment and for this great question. I wish I had the answer but it's still a mystery to me. What's odd is that the NMME models were so confident in at least a moderate La Nina as far back as late last winter, and they were confidently wrong. This seems to be distinct from the spring predictability barrier because the models were saying this event was predictable even before the spring. 

One factor that may be at play is that we have seen some big differences in the observation-based SST products (you can see that here). I don't know why we see these differences, but it leads to the possibility that some models are being initialized too cold. To support this, we have seen some pretty big Nino-3.4 errors even at very short leads. 

I'm sure there are other factors to consider, but we have to dig in to find them.