DialogFlow intents with different contexts having same training phrases

[anndtran]

Hi, 

From the DialogFlow best practice pack, I understand that generally intents should not have similar training phrases as it will affect the model performance. However, I'm keen to know what if the intents have different input contexts? Would similar training phrases still impact the model performance in this case? 

If it doesn't negatively impact the model, is it correct to assume that the model will first evaluate the intent where input contexts are matched ?

Thanks