And even with that upcoming raidz expansion feature it would be better to destroy and recreate the raidz with more disks, as thats the only way how to decrease the parity loss and increase the performance.
Example:
3 disk raidz1: 33% parity loss, 2x throughput performance, 1x IOPS performance
3 disk raidz1 expanded to 7 disk raidz1: still 33% parity loss, 2x throughput performance, 1x IOPS performance
7 disk raidz1 created from scratch: 14% parity loss, 6x throughput performance, 1x IOPS performance
So with this example you would tripple the performance and half the wasted capacity by creating a bigger raidz from scratch instead of expanding it.
As raid never replaces a backup its always good to have a copy of all your data on another ZFS pool. And in that case its not a big problem to just destroy a pool to create a bigger one later, as its easy to copy all the contents using replication ("zfs send | zfs recv"). Only the downtime while copying the data might be a big problem, depending on the size of the pool.
