So you figure if you were to use, say, a 5 SSD RAIDZ1 vdev instead of a 2 SSD mirror, that you'd mitigate fragmentation because you end up with 4x the storage space? There could be some legitimacy to that - ifff you're not tempted into using the extra space. But the keen observer might also note that you're using 150% more SSD's to accomplish that. Plus I don't think it'll pan out as well as you'd like, since there are various issues with space allocation of small blocks with RAIDZn that you don't get with mirroring. If it were me, and I had a choice between configuring ten 500GB SSD's as two vdevs of 5 drive RAIDZ1 (4TB) or five vdevs of 2 drive mirrors (2.5TB) I think I'd be tempted to go for the 2 drive mirrors and fill it to a higher percentage, because my suspicion is I'd still get some speed benefits out of avoiding RAIDZ and having more vdevs.
That might not hold true for non-block-access pools, at least not as much, but I'd suggest being prepared to do some serious testing prior to making any decisions...