Deviating from recommended defined procedures is typically a bad thing. Such deployments will inherently result in GTAC escalations and result eventually in customer dissatisfaction.
As Kurt mentions, if you intend to use all three interfaces you must follow the recommended deployment configuration and use different vSwitches. With current images, the second esa port (ESA1) will NOT work if connected to the same vSwitch as esa0. We are addressing this constraint in 9.15.
However, there are additional reasons why separate vSwitches per interface will still be the recommended deployment:
1) Bandwidth capacity - Sharing a single vSwitch is typically is serviced through a single (active NIC) which may be further shared with other load on the machine. This additional load will significantly reduce the amount of bandwidth available to the controller and could give rise to interface saturation significantly below the levels defined for the platform. Even though controllers are leased as a virtual appliance it should still be considered a switching appliance, with dedicated ingress/egress ports per platform certification. Treating it like a general virtual server can lead to un-expected performance constraints.
2) Security - Separating the vSwitch ensures that traffic from users does not inadvertently get mingled with other network traffic - post de-capsulation if the controller is supporting tunneled topologies. The separate vSwitch guarantees separation and therefore improves network security.
Therefore, for the reasons above, customers/Partners need to be strongly encouraged to follow the standard deployment configuration