A production ML model that has been meeting targets suddenly shows degraded performance and falls below thresholds. What is a likely cause of this sudden performance drop?

Drift in production data distribution is the most likely cause. When the characteristics of the data the model encounters in production change significantly from the data it was trained on, the model's predictions become less accurate, leading to degraded performance. This is a common issue in real-world ML systems. Lack of training data would typically result in poor performance from the outset, not a sudden drop after meeting targets. Compute resource constraints might slow down inference but wouldn't directly cause a sudden drop in model accuracy. Model overfitting would also likely manifest as poor generalization from the start, or a gradual decline, rather than a sudden, sharp drop after a period of good performance.