A development team is training a single, large neural network to perform text summarization for ten different languages. After a prolonged training period, they observe that the model's performance on high-resource languages like English and German is excellent and still improving. However, its performance on a lower-resource language, Finnish, has started to degrade from its previously achieved peak. The team is debating whether to continue training to further boost the English and German scores or to stop training to prevent further degradation for Finnish. Critically evaluate the long-term consequences of both proposed actions. Then, propose and justify a more effective third strategy the team could implement to address this specific performance imbalance.