When Shortcuts Return: Post-Restoration Robustness in Graph Classification

Temporarily suppressing a known shortcut can improve robustness to distribution shift, but an endpoint evaluation does not reveal how that robustness changes once the shortcut returns. We study this on a controlled graph-classification task by permuting node colors during a training prefix and then restoring the original inputs. Permanent permutation is the stronger accuracy baseline ($0.995$ vs. $0.983$ worst-environment). A trajectory experiment with ten independent data seeds and full checkpoint forks then measures the restoration horizon directly: every released run starts above $0.99$ at restoration, 36 of 40 meet a sustained-degradation criterion within 6,000 restored-input updates, and continuous suppression exceeds the released arms by $0.55$–$0.76$ anti-environment accuracy (seed-paired intervals excluding zero). Batch size modulates re-acquisition at matched update count; a longer prefix helps only at the larger batch. Norm-matched optimizer forks show the reset damage is largely a scale-shock artifact. Diverse training mixtures support robust ERM without suppression, environment grouping changes the effect of an invariance penalty, and an external task with naturally diverse shortcut proportions reverses the ordering—bounding when suppression is the right tool. We distinguish these observations from unresolved mechanisms and give conditional analysis for low CE gradients and fixed-feature replacement.