Hi,
I've experienced strangeness with boot-from-SAN in the past, not just with ESOS arrays, but with other vendors' Fibre Channel disk arrays as well. I don't have any experience personally with Emulex FC HBAs on the initiator side, but have used QLogic in the past. One of the issues was it seemed as though the adapter or possibly the fibre channel switch "cached" a previous configuration and in the LUN screen in BIOS it would show a previous (no longer mapped) volume. Seems like some full power-offs on the initiator side cleared that up.
I wonder if the Emulex FC BIOS is looking at the ALUA configuration at all, or doing anything with that. Also, if I remember correctly from my SAN boot days, if I had multiple HBAs (adapters) on the initiator side, and the volume wasn't accessible on all paths, it would not boot too. With the Emulex BIOS do you specify the order of paths to try? Maybe its just trying the first path? Or randomly selecting one of the paths if multiple are configured? Some of your paths will be advertised as active/non-optimized so maybe its trying those non-optimized paths? Maybe it ignores ALUA all together, I'm honestly not sure.
I'd start with disabling the ALUA configuration... maybe take down one node so it doesn't make your cluster freak out, copy your /etc/scst.conf file and then remove the ALUA parts (device groups, target groups) and start with that.
--Marc