Hugh performance difference when calling RecommendationServiceClient.apply_recommendation with request vs with customer_id and operations

[puerile22]

When I tried to apply recommendations by using ApplyRecommendationRequest, it takes around 5 minutes to make 100 operations, but if I do it with something like this:

recommendation_response = recommendation_service.apply_recommendation(
        customer_id=customer_id, operations=operations
)

It usually finishes within seconds, the operations list also contains 100 apply recommendation operations.

This is how I build and apply the apply recommendation request:

request = client.get_type('ApplyRecommendationRequest')
request.customer_id = customer_id
request.partial_failure = True
request.operations = operations

recommendation_response = recommendation_service.apply_recommendation(
        request=request
)

Is it because of the partial_failure flag? Why is there a huge performance difference between these 2 approaches? Also when I use ApplyRecommendationRequest approach, I usually get a bunch of errors like Authorization of the client failed.

I'm using the latest version of this library.

Any help would be much appreciated Thank you.