I don't care what you're app is. Properly configured, you can stand up a duplicate machine effectively immediately with a backup of the virtual disk images and/or dedicated HD(s).
That's one of the bullet points on every VM sales sheet.
You've moved the goal posts a couple of times but your last post is more correct. An regular capture of the image itself would allow the system to be restored more easily. There would probably still be a fair amount of work to reintegrate it back in to the domain, especially after it being down a couple of days.
And come on guys, I'm sure Oyeve doesn't need the pounding he's taking in here. What's done is done. Time to move on.
My advice would echo drebo. How big is the organization? You can migrate your users to a hosted solution fairly cheaply these days. This way you don't worry about downtime or backups.
Now, if it has to be a self hosted solution, for whatever reasons (security/compliance/whatever), then yes, grab the offline cache as soon as possible. At that point, I would put a support call in to Microsoft, explain what happened and tell them you want to remove all traces of the e2k3 server from your domain. This will cost you about $260, but I don't believe anyone can do this as well as Microsoft can, plus they really support you for just that one payment. If they say they did it and two days later you run in to problems, they'll support you. Broaden the scope by saying you had a 2k3 server die, you can't restore it, and you need to replace it with a 2k10 server. That way the installation of the 2k10 server is included in the scope and the support call cost. They'll basically fix any issues you run in to.
And use dedicated hardware for exchange. I don't believe it works great in a VM to begin with. Smaller organizations could get away with that much easier though.