might be feasible, technically ... but sounds complicated and definitely has lot of pitfalls, which could only be handled with lot of NetApp-experience.
regarding complicated: you probably need first to move away the data from one controller, then move it to cluster, then rebalance it once the cluster is finished, so you move your data around and around ...
regarding pitfalls: if HA is single chassis you need to manipulate chassis configuration. Also I don't think that it is even possible to set up cDOT in single node cluster mode when second controller is in the chassis. And even if: albeit there are how-tos to get from single-node to "normal" HA pair cluster configuration, it means additional work. Same for dual-chassis HA configuration.
So why not following NetApps standard procedure: get some swing gear, build up a new cluster with the 4TB disks, move the data, destroy the existing FAS3240 config and add them to the cluster (keep in mind you need disks for root-aggrs), remove the swing gear. No pain with HA/non-HA config, no single-node/multi-node reconfiguration, less data movement.
Re-use the disks from the old 32-bit aggrs if appropriate.