I can see that stratified sampling helps in maintaining the same class distribution in the training set as in the original dataset. However, my understanding is that ideally, the model should be trained on an equal number of instances from all classes and if the class distribution in the training set is itself biased, the resulting model would also be inherently biased. Is this correct?
It is typically not ideal for a model to be trained on 'an equal number of instances from each class', unless that is representative of the real population. See Are unbalanced datasets problematic, and (how) does oversampling (purport to) help?. The point of stratified random sampling isn't to remove class imbalance, because, as you point out, it does not do that. As for the actual benefits, see Benefits of stratified vs random sampling for generating training data in classification