I thin I'd prefer a cold spare to a hot spare. There's probably more change of failure happening at around the same time if a pair have been running as replicas or one as a hot spare. A 15 min invoicing downtime while you unbox and boot the cold spare and restore from the usb stick from the failed one - is probably not as big a deal as the risk of one of your replication pair going bad and possibly corrupting the other, or having the similar runtime secondary fail around the same time as the primary.