Using SparkTrials for parallelizing hyperparameter tuning can lead to issues such as resource contention and communication overhead. These challenges can be addressed through careful configuration and monitoring. However, there may be scenarios where SparkTrials is not the best choice, such as when the model is too small to benefit from parallelization or when the overhead of setting up a Spark cluster outweighs the benefits.