1) “A sound Exchange Server in a sound Active Directory and DNS infrastructure”, that’s what I say. Since 95% of all Exchange Server errors are DNS and/or Active Directory related (statistics provided by me, so they are as skewed, partial and invalid as can be), you first have to keep your AD and DNS running smoothly.
2) Never let an Exchange Mailbox Server handle more than a 1,000 users. You’ll thank me later. Or, now. Now is a good time, actually 🙂 OK, there is hardware that can enable Exchange Server to handle more than 1,000 users, but are you sure you have it? Will you be able to perform backups easily? Will they finish quickly? Will you able to restore everything super easily and super quickly? If not, then each server should handle 1,000 users at most. At most! 🙂
3) Documentation, documentation, documentation. Document everything. Every procedure, every change, every event. Obsessive Compulsive Disorder is a nice trait for an Exchange administrator to have 🙂 You also have to be able to rebuild the whole infrastructure at a moment’s notice, should the need arise. Therefore, you should also document the disaster recovery procedure as clearly as possible. You have to be confident and to be able to prove your confidence to yourself and to upper management. Demonstrate full restores to dissimilar hardware at a different network, either in a lab or a disaster recovery site.