No reason I can think of why you can't run two of them in a Linux software RAID-1 set.
What I wonder about, is whether SSDs of any sort can be trusted to report data that's gone bad in the same way as hard drives can. Checksums and ECC codes are utterly fundamental to spinning rust drives, but even with these it's possible for a controller failure to allow undetected data corruption. And RAID-1 normally assumes that if a write succeeded without error, a read may be satisfied with data off either disk in the RAID array without checking that it's (still) the same on the other device.
For the paranoid, these devices might be fast enough that a new RAID class could be defined and used without crippling loss of performance. Minimum three mirrored members. On read, get the data from all three members and check for equality. If one differs from the other two, it can be assumed to be the bad one. Two members wouldn't let you know which was the good one (assuming that the bad drive wasn't detecting its own failure state).
Like navigating with one chronometer or three chronometers, but never two.