What are your fault tolerance policies if a minirig failed? I assume that one minirig could hash at a rate of 1500GH/s (as per BFL's advertising) and assuming that 1GH equates to a minimum of one contract per GH, then 1500 contracts would be affected.

BFL have not issued the specifications of the minirig and I am assuming it has no fault tolerant hardware or High Availability hardware either; such as hot swappable PSU's or ASIC racks if there was a failure within the minirig. If this was the case, perhaps you should look into a fail over cluster to deliver the hashing capacity as promised by your contracts.