I'm inclined to agree with your boss. A terminal server doesn't need fast disks, which RAID 0 would enhance. Your I/O bottleneck is going to be your network.
RAID 0 configuration, which is striping, doubles your risk of disk failure. You might want to consider RAID 1 which is mirroring.
SCSI disks are expensive. Here again we run up against the issue of disk performance and how it is not an issue in your terminal server. I'd use ATAPI disks. I'd use Seagate or Maxtor, and not the expensive server disks either. Desktop disks will perform fine. I'd avoid Western Digital disks.
SCSI costs more money in its required support hardware. Either you have a motherboard with a SCSI controller, which is more expensive than a motherboard without a SCSI controller, or you need to purchase a SCSI bus card. So you can save money both on the disks and on the controller if you purchase ATAPI disks and use software based RAID 1.
Think about this. The slowest, cheapest CPUs that you can purchase these days are overkill for this application. This machine will not be busy. The network cards may be saturated but the CPUs may still be 50% idle unless you are going to have six or eight network cards. The workload on this machine will be very bursty rather than evenly spread out over time.
I think that you can use good, solid, but inexpensive hardware to implement this machine. You can get reliable performance requirements from your other terminal servers so that could alter almost anything that I said about CPU requirements, but the disks still probably don't have to be high performance server disks.