Question 1 - I would definitely recommend rebuilding your HA pair using ONTAP 8.3 or newer (9.0 is GA and 9.1 is pending very soon). I say rebuild as there is no way to move from 7-mode to Clustered ONTAP in-place, so you'll need to move any data you need off first and move it back (if you need temporary storage your NetApp rep/reseller can probably help you with temporary "swing gear" to get it done). Then, using ADP you can have a single large aggregate using 22 or 23 of the disks (2 or 1 spare) and also gain the performance of the extra spindles. Do bear in mind that you would be making the 2 nodes active/passive, as there would be no data access on the opposite node of the data aggregate outside of failover scenarios, but you wouldn't be able to overwork the CPU and memory of the single FAS2552 node with the small number of disks anyway. If you later add more shelves, you can either add the disks to the same aggregate or create a new one on the other node and spead the workload to make it active/active.
Question 2 - This is correct, in 7-mode you have 2 paths from your host(s) to the first controller, and if you lose both of those paths your host(s) lose access. Using Clustered ONTAP, you would have 4 iSCSI paths from the host(s) to the SVM (Storage Virtual Machine), as only the disk aggregate(s) are directly linked to the individual nodes of the cluster. This also means if you have multiple aggregates, you can safely move LUNs and Volumes between them with no outage - the SVM simply moves it's back-end access to the data when you move between aggregates. This is a big win in larger enviroments, and can also mean no outage when you do storage system refreshes - you simply add new nodes to the cluster, perform Volume Moves, then retire the old nodes (of course I left out some details but you get the drift). The SVM is an abstract above the physical hardware, just like using a hypervisor for compute resources. 7-mode HA will failover the entire ONTAP stack onto the opposite node (which can be manually invoked using cf takeover commands or by ONTAP seeing hardware failures and auto-failing), which does make the system highly available from a hardware failure or software upgrade view, but Clustered ONTAP (now simply known as ONTAP again in 9.0+ since 7-mode is depreciated) is much more more capable of making network interfaces, protocols and your data highly available (and also up to 24 nodes, not just 2).
I hope my long-winded explanation makes sense and helps, let me know if you need more info.
