Hi,
1) Yes, Multicore does benefit the throughput. This might not be visible in a single throughput lab test, but in a larger network, definitely yes.
2) Yes, sending data means compression and dedup, which is a lot more processing intensive than decompression and reduplication. Hence, the 20 Mbps upload would likely be the bottleneck at the remote sites. The VM route can work if the storage is local.
400 Mbps is a tall order, but it is possible with the three remote sites. I take it the Internet breakout makes up for the remaining 100 Mbps. If correct then optimized traffic is closer to 300 Mbps which makes it a bit easier.
The static settings of Express might just just work, if the setup is perfect. If the POC goes well and you want some guarantees then I would recommend going with the Plus route so that we can work with you to ensure the 300-400 Mbps is achieved reliably. It also enables a larger datastore which will be very useful at these throughput speeds.