There's several issues that come to mind. The first and most important is parallelizing the storage: Whenever possible, it helps to place each virtualized disk on its own physical spindle to reduce contention. Also, depending on the amount of network traffic to and from these SQL boxes, you may want to try and dedicate a single physical network interface to each VM whenever possible. If you think you can get away with putting everything on a single gigabit Ethernet connection, try it -- just be mindful of what you're doing.
Note that if these are machines that don't get a lot of traffic or utilization and you're merging them mostly for the sake of legacy support (i.e., they're going to be phased out or discontinued at some point), you can often get away with more resource sharing (e.g., multiple virtual discs on one physical disc), but again, the mileage will vary heavily depending on the actual usage.