I'm fairly sure this is all correct... You can add extra drives to the ZFS "pool" but they can't become part of an existing RAIDZ array (unless as you pointed out they are replacing an existing smaller disk). So for instance if you had 4*1Tb (~3Tb available as storage) in a RAIDZ array you could add a 5th 1Tb disk to the ZFS pool and your available storage on the pool would go up to 4Tb, but the extra 1Tb you added would be just an individual drive and a failure of that new drive would not be protected by the redundancy of the other array. (Correct me if I'm wrong guys, I've been trying to research this myself also!)
What you could do for example is have 3x1Tb drives in a RAIDZ1 (RAID5) and then at a later stage add another 3x1Tb drives in a second RAIDZ1 which would give you a pool with 4Tb storage overall and something similar to a RAIDZ2 (RAID6) level of redundancy (though not quite, because you could only tolerate 2 simultaneous failures if they weren't on the same array).
While we're on the subject of ZFS, I've been reading around about using a ZFS based system within an esxi VM and there are quite a few places I've seen people giving pretty serious warnings about doing so... Mostly in the sense of "don't do this unless you know exactly what you're doing", "many people have lost all their data by doing this", etc. etc. and a lot of discussion suggesting you have to use RDM to get a NAS VM working correctly (and whether this is a good idea or not). Can anyone help clarify things a little here?