Questions about FirstSpot Implementation

[GoldChain]

I was looking at the following page:

http://www.patronsoft.com/firstspot/topologies.html

I believe I would be interested in implementing Scenario #3. More specifically implementing that scenario using VPN tunnels between sites as seen here:

http://www.patronsoft.com/firstspot/FirstSpot_Scenario_3_with_VPN.pdf

However, I have a couple of questions:

1) How does the unecrypted tunnel traffic, exiting at the central site, get directed to the FirstSpot server instead of going directly to the local router to be routed to the Internet? A static route on the tunnel device? If so, doesn't that mean that *all* outgoing (toward the internet) traffic will be directed through the FirstSpot server (I realize that return traffic would go directly back to the client, so it's wouldn't really be considered a proxy, but at least all outgoing traffic would go thru the FirstSpot box?)

2) Understandably, when a FirstSpot server is at each "hotspot" location (as in Scenario 2 from the first URL above), the server is able to recognize new connections based on MAC address. However, in the scenario I'm asking about, since the traffic crosses Layer 3 boundaries, does the centralized FirstSpot server simply use IP addresses to recognize new sessions?

3) On a related topic, what is a good way to guage how many simulateous users could be serviced by a single FirstSpot server? (I realize that it's probably based on the hardware capabilities on the server, but a baseline would be good to know for sizing a server to purchase)

Finally,
4) Since a single point of failure is not a good thing, does FirstSpot implement any type of redundancy so that if two boxes were purchased they could either be setup as active/passive failover or even load-balanced?

Thanks for your input.

Mike W.