How about this: You need to buy 4 drives. Buy four identical (good ones), and run RAID-Z2. Now you have two-fault tolerance. If one drive fails, just run in degraded mode for a few days (but you have not lost data, and are actually still 1-fault tolerant), while you order a replacement drive. Once the replacement drive arrives, put it in, resilver, and you are back to full performance.
Why do I say this? Because I don't like spare drives. They sit there and do nothing; their return on investment is zero most of the time (except in the rare case that you actually need them). If you have to have the 4th drive anyway, then make it do something useful: let it give you extra redundancy. Actually, there is an ulterior motive behind this. Remember, with modern drive sizes, the probability of finding a latent fault while doing a full resilver is very high (it is approaching ~1 for 10TB drives). So using RAID which is single-fault tolerant is a really bad idea, because once a drive completely fails and you have no fault tolerance left, the probability of data loss is very high. So much so that the former (now retired) CTO of NetApp has referred to 1-fault tolerant RAID as "professional malpractice". By using a 2-fault tolerant code you avoid this problem. And by having no explicit spare, you use a nice modern technology called "fail in place": Let the drive fail, deal with it failing, and replace it at your leisure, not in a big panic, nor by having a hot spare on standby.
