The two servers just co-exist. They dont "act like they are one organization". I dont know whether you ever managed a non-trivial exchange setup, but exchange can habe multiple servers in one organization and often does so as part of normal operations - for example when you have significant offices in many locations, or jsut because you dont want downtime and happen to have 100.000 mailboxes which are too many for a server to handle anyway - so you cluster them.
For example Microsoft - it makes a lot of sense to have servers in germany (even in different larger locations), and not run everything through one monster system in Redmond. Simply because there are a couple of thousand people here in germany working, enoug hto actually fill up some redundant servers.
This is actually very nice as you can move peoples mailboxes between locations if they move (not for a business trip, but when you send a guy a year to your indian outsourcer - move his mailboxes there, too), and with 2010 you can have mailbox replicas on backup servers in case your primary server goes down, all without a lot of administrator intervention.
2007 / 2010 coexistence basically means you can introduce new 2010 servers, then move all items of the 2007 servers. You do not have to upgrade all your servers at the same time.
Which server handles the mailflow? Hm - makes me wonder. Ever read the documentation? Mail flow is handled by all servers that are in the corresponding role (Hub Transport Role). Exchange does a lot of optimization under the hood to autoamte that - the real administrative beauty compared to most non-scaling and primitive linux mail daemons here. Although you need to still set up connectors (to external endpoints) and assign them to hub servers (which makes sense, for firewall routing, where you need to know where exactly internal email may come from). All this is - nitpicking - very well described in the introductory chapters of the documentation. You know - concepts and such.
What really is the overall point? Well, an organization like Microsoft has possibly a couple of dozen mail servers around (not to say some hundred), and upgrading them all at the same time would be hard, and still interrupt business. Now, in your world you probably get paid regardless whether your company makes money, but the company has a strong interest NOT to let 100.000 of their emploees get out of communication for possibly some day or two without any reason. That is the whole point. Business continuity. Maintaining uptime. Being in existence. And not having to upgrade servers on the whole blogis at the same time.
