Sorry to take so long to get back to this.

Yes, we consolidated a lot of _production_ servers/services into the two ESX boxes. Yes, that puts a lot of eggs in the two baskets. That's why you start with really, really good hardware with top drawer, 4-hour on-site, send the tech, quick-turnaround warranties. We bought HP 580s with the best warranty. With the exception of a couple hicks early on, no problems for almost 18 months ... and still cooking.

The DCs and DNS are still on separate, hard boxes. No virtuals for those services.

Also, with two boxes, there is enough room to move the virtual servers from one to the other while you take the one down for repair, maintenance, updates or whatever. This can be accomplished without disrupting services in any user-noticable way through Virtual Center. Things get a little slow for a while, but it works.

Another thing we did. We set up two physical-virtual clusters. One for Exchange and one for SQL. Each cluster had one physical box (active node) and one virtual machine (passive node). This was cool because you really need the hard systems for Exchange and SQL on the active node. But the passive node only needs to be a virtual system. Failover causes a slowness for a while, but doesn't kill the systems.