iscsi ... at some point zvols and the networking info used to share them out adds significantly to fail over times.
if you use extensive snapshotting this can also dramatically increase export/import times. imagine having 100 file systems you share over NFS, each file system has 100 snapshots ... importing can take awhile.
i haven't yet had a split brain. i know they 'can' occur but from what i understand it is extremely rare.
as for routine maintenance in general it is often best to do a reboot -p or udamin 5 0. this forcibly halts the system and as quickly as possible hands control to the secondary node.
just as an example from my testing. my volumes have 120 data disks (240 per cluster). on average it takes roughly 40 seconds to do a full manual export/import. nexenta 4.0.x has some improvements which decrease those times. I haven't tested them yet but some of the export/import functions used to be serial in nature and from what i understand are now parallel. unfortunately iscsi/comstar stuff still (to the best of my knowledge) isn't parallel so it still takes awhile with lots of zvols etc.
forcing a system halt typically cuts out about 15 seconds.