The fact that we're continually exceeding the most pessimistic projections the models can produce on a yearly basis, even after repeatedly recalibrating it to build in even more pessimistic assumptions, and that we're tending to depart more strongly from even the pessimistic models.
This implies that the models are broadly understating these feedback loops, that they are net-balance in the positive-feedback direction, and that they are becoming more active in the current climate behavior regime.
If we continually exceed model predictions, it could be from underestimating known positive forcings, overestimating known negative forcings, or some combination of unknown positive/negative forcings.
Essentially I think continuous failed predictions are a red flag for any model, especially one as critical to us as the global climate models.
Strictly speaking that is not sufficient as it only holds for the current part of the climate evolution and not the future one. It could be that currently positive forcing is underestimated while in the future negative forcing is.
