You can still hot-remove even if it is not supported. I've been doing it for a while.
1. Destroy volumes on the aggregate on the shelf .
2. Destroy aggregate on the shelf.
3. Remove ownership of the disks (priv set diag ; disk remove_ownership -f , etc)
4. Move the cabling from the head, to your IOM modules, one at a time, to the next shelf down, in the same place (circle, square, etc).
5. After each cable move, wait and watch the console of your filers to make sure they don't die.
6. Move the next one, and so on.
You'll get an autosupport email about loss of multipathing on your filers, and you'll get an autosupport about the heads no longer having valid paths to specific disks, but it'll stay up, and continue just fine. When you finish your cable moves, both heads should tell you that multipathing has been restored. storage show disk -p should show multiple paths to all shelves and disks (save shelf 1, which you removed.)
Again, I've been doing this for a while, on multiple kinds of heads and in multiple clusters - but all 7mode, mind you. I haven't done this on cluster mode yet.