How to validate point process predictions or can tests for calibration be proper?

In this talk we consider probabilistic forecasts of point processes and discuss how they can be validated obeying decision-theoretic principles.
In practice, such forecasts are often assessed by testing whether observed point patterns could be random draws of the predictive model.
We ask the question whether such a test can be proper, meaning that the expected outcome is best when the forecaster predicts the true distribution.