Just wanted to chime in on this. I'm having similar problems with the drives detecting as different numbers.
I'm running this in esxi 5.5 with 4x 3tb drives on xpen 4. raw disk mappings work just fine, but my goal was to pass through an ibm m1015 (LSI9211-IT) to the guest. This way I would get smart data and temp readings.
My issue seems to manifest with the power on/off of the drives. Basically if the host is off and I boot it, all the drives spin up during the host boot. If I then start the xpen VM everything works and is detected fine. In this case the disks are already spinning because they spun up with the host. If I then shut down or reboot the xpen VM, I can hear the drives spin down. When I start the VM again, I can hear the drives do a staggered spin up. After that, the drives end up in all sorts of order in the GUI. Sometimes one will be 12, next boot it'll be 20, but they are all always detected. Reboot the host, and I get another good boot with drives detected correctly.
What I think is happening is that the drives aren't really ready when the is is going through the drive detection. If I go into the GUI I can watch as the drives detect well after the GUI has loaded. The first disk it detects is always initialized with a crashed raid warning, and the other 3 are not initialized.
My thought was to try adding a boot delay to grub to try and give it time to spin up the drives before it starts the actual boot process. Any thoughts on the likelyhood that may do something?
Thanks