I've been recently experimenting with converting transistor-level SPICE models of ICs into subcircuits. One thing I try to keep consistent with my subcircuits is that they remain floating; i.e. no internal connections to node 0 (aka GND). This represents the real world more accurately, and also allows for inserting noise sources or series resistances into the ground pin on the subcircuit. Any internal connections to node 0 would bypass these external effects. I've successfully made behavioral opamp subcircuits which easily satisfy this.
Then, I moved over to trying my hand at an LM741 since its internal transistor-level schematic is available in the datasheet (page 7). When I parsed the netlist into a text file to make the subcircuit, I was reminded that all BJT substrate nodes in SPICE are by default connected to node 0. If I was making a logic IC or regulator IC subcircuit, I would simply remap the node 0 connections to whatever the ground pin on the subcircuit is. However, this is an opamp and has no ground pin. Just VCC and VEE supplies.
So this begs the question...what are the BJT substrates in a bipolar-based opamp IC connected to? Are they connected to VEE? Are they left floating? Are they shorted to each individual emitter? I know next to nothing about IC design, and this situation piqued my curiosity.