Yes - even if you are only allowed look at the validation set twice a day, you can still overfit it, with sufficient days.

But thats hardly a criticism of Kaggle.

Its the responsibility of the person building the ML model to take steps to avoid overfitting. Even if competitors fail to take these steps, the fact that Kaggle has a built in validation-set helps avoid overfitting.

Yes, with enough time it would still be possible to overfit, but I don't see how that is Kaggle's fault.

If anything, they've a framework to discourage overfitting, not encourage it.