http://www.vmware.com/pdf/vsphere4/r40/vsp_40_availability.pdf pg 34 which you have probably found already.
I don't really understand what you mean by b), but the idea is you are tying up memory/cpu from 2 different hosts pointing at the same vm on a datastore. If that vm were to move, then both hosts would have to have their datastore pointer updated at the same time and they chose to not make this possible. Think of the checks that both hosts need access to the same shared storage that the VM would be moving to, it would be extra checks the svmotion process would have verify.
However, as it indicates the workaround is trivial. Disable FT, svmotion your vm and turn it back on.
NCIE, VCAP-DCA[45] #283, VCP5-DT - http://sev3.net - Severity 3