Does sharing a common GlobalPooling layer for different Convolution outputs have implications on gradient flow?

Question

I am building a CNN and I am not sure if I can share a GlobalPooling layer across diferent outputs of previous Convolutional layers. In guess yes,