To start, understand the physical hardware topology inside and outside of the host server. Network adapter or interconnect cables can fail, as can a networking switch port or even the switch itself. The failure can be much more localized. If a dual-port network adapter is a mezzanine card attached to the mainboard of the server, and the mezzanine card suffers a failure, both of the network adapter ports can experience an outage at the same time. As a result, plan for failure with adapter replacement. Multiple network adapters with different host backplane connections can help minimize the impact of an adapter or port failure.
Workloads can also saturate one or more interconnects. Even mundane tasks such as server or application-level backups can saturate a network and cause cascading impacts to other services that depend on that interconnect path. Depending on your server architecture, multiple networks can help minimize the impact of these operations. Separating purposes such as server and host management, CSV connectivity, Live Migration communication, and backups could be better suited with multiple network adapters specific to a single purpose.
In some cases, having multiple sets of network adapters for different purposes can be overbearing and a cable or switch port management nightmare. Limiting network visibility through either port isolation or VLAN tagging can help multiple distinct purposes coexist on the same physical network while providing logical isolation for each of these tasks. Quality of Service (QoS) can help prioritize certain types of traffic over others to minimize the impact to these workloads. VLAN tagging to segment the network traffic can be managed within the network configuration of the Hyper-V host, as long as the physical network is configured appropriately.