guardRejections

The number of times the guard rejected a proposed step across the most recent run.

This is not a debugging statistic. It measures how far the estimator catalog departs from the exact maximum likelihood estimates the monotonicity theorem assumes, so it belongs in the experimental record.