The single most important investment you can make in your infrastructure has nothing to do with NServiceBus at all, and everything to do with hardware virtualization.
The whole point of NServiceBus is to provide transactional messaging so that you don't lose data. This is moot if you host NServiceBus on a physical server that could go up in smoke at any moment, taking its messages with it. With a properly configured virtualized environment, this risk basically disappears. In the event of a hardware failure, the hypervisor should be able to migrate the virtual machine image to another host, often completely transparently, where it will resume processing messages without missing a beat.
Besides hardware failures, driver problems are the other main cause of catastrophic system failures. In most cases, virtualization removes this problem as well because the drivers are abstracted, generic drivers managed by the virtualization platform in order to support hosting the guest system...