Every FlowNode has its own ID value & descriptorId which is a separate instance – however since IDs are currently numeric and thus have a limited set of values, and descriptors have a limited set of values, this just wastes memory.

We should be interning these values (or similar).