Considering the Edge Role ( Active-Passive) , you can certainly create in that way and take care of DRS rule (Manual process) . By any chance we have a stretched storage layer ? While placement of edges are very key design decision , it is equally important to understand failure scenarios like Active Edge fails and Passive takes over the role . You have mentioned DCI is 10gig pipeline , so I'm assuming for DC-DC traffic for whatever reasons and data traffic for NSX workloads will be sufficient enough even during a failure scenarios (When VM's are on one site and Edges are on another site) and irrespective of the failures , there will be traffic hair pinning. Redundant DC's is a good approach but do understand the nature of the traffic flow this will help you align closely with missing gaps.