Yes you need mirrored if you don't want issues to appear/lower system uptime. I'm not using an SSD for storage - I've seen many SSDs die at work in workstations and SLC-based SSDs in servers in under 1-year of duty. Platter drive failures are more common than they should be as well - my older 2TB Samsungs are still the most reliable I have and have outlived the 3 and 5TB drives I have owned. Newer drives, many test bad out of the box, including more expensive WD Blacks and Reds. Seems we have about a 15% rate of failure on initial 1st-pass butterfly testing when we receive new-in-box drives (some SAS, some SATA).
I had one 6TB WD purple drive. It died in 2-months. Anecdotal there as it was a one-off personal purchase, not a purchase of 50+ drives like we do at work. Wasn't a great start to a newer line from WD.
I'll stick with SSD for read-heavy loads and mechanical HDD for write-heavy loads for a few more years.
Archival storage is on a separate disk-array server, backed up to a NAS and another external USB drive. I'm not too concerned with small 500GB drives for initial-clip storage in the BI system. if we can get shadow-motion under control, 500GB would store even longer duration.
I agree SFF is a great overall size for these things though. Some only have two or three SATA ports. I specifically look for at least 4. Power usage also isn't an issue - if you limit a desktop CPU to a lower speed and undervolt, you will also see power consumption close to the numbers from a NUC. Having not done any testing recently, years ago I had a Core2Duo E8400 system that I ran at 1Ghz - much faster than an Atom and peaked at around 15W usage. Atom idled at a couple Watts less, but we won't see idle/sleep state in a surveillance system. Hardware price was also about the same - the Atoms were overpriced.