My opinion on this is using mixed designs, I guess 10G/40G/100G you are utilizing the impacted band, which has lower performance can be used for 10G traffic for shorter distance. Whereas, band with higher performance may be used for higher data rates for long distance. Therefore, this may improves the supported capacity for the service providers. Hence, revenue per unit of invested cost is also increased. Cost may be any thing may be power or energy units. Please let me know your view about this.

I agree partially. In my opinion, what you have stated is true provided the sub-bands (spacing between bands of same rate and/or spacing between bands of different rates and modulation format) remains constant. Once the sub-bands are varying, performance of the bands also varies. This will impact the QoT (BER) and cost. Of course by cost I meant power consumption. Please let me know your view(s).

Agree with this. Moreover, this wise variations in the spacings can helps to deco-relate the impact of Nonlinear effects and it can be used in the positive way to boost the OSNR of the services. 

In my view, anything we achieve using simulations should have the feasibility with the field deployments as well. This variation in the spacings should be supported with the switching equipments in the field.

Yes, agree with you view. The switching will play a crucial role. I am thinking of how to model a network that can provide high spectral efficiency and low power consumption (both at the same time) for varied BER which will depend on the demands that come into the network.